1 00:00:00,080 --> 00:00:02,500 The following content is provided under a Creative 2 00:00:02,500 --> 00:00:04,019 Commons license. 3 00:00:04,019 --> 00:00:06,360 Your support will help MIT OpenCourseWare 4 00:00:06,360 --> 00:00:10,730 continue to offer high quality educational resources for free. 5 00:00:10,730 --> 00:00:13,340 To make a donation or view additional materials 6 00:00:13,340 --> 00:00:17,217 from hundreds of MIT courses, visit MIT OpenCourseWare 7 00:00:17,217 --> 00:00:17,842 at ocw.mit.edu. 8 00:00:21,012 --> 00:00:22,470 PROFESSOR: Today what we want to do 9 00:00:22,470 --> 00:00:25,750 is use the two papers that you read as kind of a backdrop 10 00:00:25,750 --> 00:00:29,660 to try to think something about the regulation of genes 11 00:00:29,660 --> 00:00:31,730 in response to changing environments. 12 00:00:31,730 --> 00:00:34,130 So there's the Mitchell paper that 13 00:00:34,130 --> 00:00:37,880 is talking about this idea of anticipatory regulation, 14 00:00:37,880 --> 00:00:40,530 whereby if the environmental changes have 15 00:00:40,530 --> 00:00:43,330 some typical pattern, then maybe the cells can take advantage 16 00:00:43,330 --> 00:00:46,480 of that and start preparing for environment number two 17 00:00:46,480 --> 00:00:49,000 when the cell sees environment number one. 18 00:00:49,000 --> 00:00:52,470 But in other cases, it may be that the environment fluctuates 19 00:00:52,470 --> 00:00:55,370 in ways that are really fundamentally unpredictable. 20 00:00:55,370 --> 00:00:58,900 In that case, you can't use the sort of anticipatory regulation 21 00:00:58,900 --> 00:01:02,130 strategy, but instead, there may be a way 22 00:01:02,130 --> 00:01:04,459 that you can just stochastically switch 23 00:01:04,459 --> 00:01:07,440 between the different strategies and implement what's known 24 00:01:07,440 --> 00:01:09,140 as a bet hedging strategy. 25 00:01:09,140 --> 00:01:11,490 And this is modeled largely on the work 26 00:01:11,490 --> 00:01:14,910 that you just read about in the Kussell paper, Science 27 00:01:14,910 --> 00:01:19,870 maybe May 2005. 28 00:01:19,870 --> 00:01:24,560 I do want to stress, however, that if you observe 29 00:01:24,560 --> 00:01:27,550 phenotypic heterogeneity in a clonal population, 30 00:01:27,550 --> 00:01:31,040 that does not necessarily mean that the cell or the population 31 00:01:31,040 --> 00:01:33,970 is implementing one of these bet hedging strategies. 32 00:01:33,970 --> 00:01:36,600 In particular, I would argue rather strongly 33 00:01:36,600 --> 00:01:39,256 that there are other possible evolutionary drivers 34 00:01:39,256 --> 00:01:40,630 for such phenotypic heterogeneity 35 00:01:40,630 --> 00:01:41,730 in the population. 36 00:01:41,730 --> 00:01:44,390 First of all, just because you see 37 00:01:44,390 --> 00:01:47,740 some phenomena does not mean it is necessarily selected for. 38 00:01:47,740 --> 00:01:50,830 So it's possible that it's a side effect of something else. 39 00:01:50,830 --> 00:01:53,960 However, if you're looking for an evolutionary kind 40 00:01:53,960 --> 00:01:55,857 of explanation for something like this, 41 00:01:55,857 --> 00:01:57,190 bet hedging is not the only one. 42 00:01:57,190 --> 00:01:59,030 In particular, we'll talk about two 43 00:01:59,030 --> 00:02:00,780 other possible explanations, and they both 44 00:02:00,780 --> 00:02:02,870 have to do with kind of game dynamics 45 00:02:02,870 --> 00:02:05,900 that we might illuminate or talk about. 46 00:02:05,900 --> 00:02:08,101 But I guess our game theory talk was-- 47 00:02:08,101 --> 00:02:10,100 that was a week and a half ago, almost two weeks 48 00:02:10,100 --> 00:02:12,170 ago now-- so maybe you've forgotten all the game 49 00:02:12,170 --> 00:02:13,550 theory that we discussed. 50 00:02:13,550 --> 00:02:15,400 But in particular, it could be that 51 00:02:15,400 --> 00:02:17,733 the phenotypic heterogeneity might be the implementation 52 00:02:17,733 --> 00:02:19,390 of a mixed strategy. 53 00:02:19,390 --> 00:02:22,640 Or possibly it could be an example of some sort 54 00:02:22,640 --> 00:02:24,630 of altruistic self-sacrifice. 55 00:02:24,630 --> 00:02:27,980 And we'll try to explain the theory behind each 56 00:02:27,980 --> 00:02:31,790 of these three things as well as possible biological examples 57 00:02:31,790 --> 00:02:32,920 of each of the three. 58 00:02:32,920 --> 00:02:38,240 In this case, we may think about bet hedging 59 00:02:38,240 --> 00:02:43,770 as maybe an explanation for antibiotic persistence, 60 00:02:43,770 --> 00:02:46,150 this idea that cells can switch into these slow-growing 61 00:02:46,150 --> 00:02:47,525 persister states in which they're 62 00:02:47,525 --> 00:02:49,680 resistant to antibiotics and other stresses. 63 00:02:49,680 --> 00:02:51,900 Mixed strategies could be-- well, 64 00:02:51,900 --> 00:02:54,570 we're going to argue it could be implemented 65 00:02:54,570 --> 00:02:57,080 in the context of mixed sugar environments. 66 00:02:57,080 --> 00:02:59,680 And altruistic self-sacrifice may 67 00:02:59,680 --> 00:03:03,954 be the explanation behind colicin production in bacteria, 68 00:03:03,954 --> 00:03:04,620 so it's a toxin. 69 00:03:11,060 --> 00:03:14,230 So I want to start by thinking about this thing 70 00:03:14,230 --> 00:03:17,030 about adaptive prediction of environmental changes 71 00:03:17,030 --> 00:03:18,050 by Mitchell. 72 00:03:18,050 --> 00:03:20,810 I think this is a very interesting paper 73 00:03:20,810 --> 00:03:22,700 in a number of different ways. 74 00:03:22,700 --> 00:03:26,600 One is, I think that it's sort of a big idea that 75 00:03:26,600 --> 00:03:29,550 can be explored in these simple experiments. 76 00:03:29,550 --> 00:03:31,990 I think that it's an exceptionally clear paper 77 00:03:31,990 --> 00:03:34,870 in some ways, and that they really say, 78 00:03:34,870 --> 00:03:37,300 oh, we're going to propose that this strategy should 79 00:03:37,300 --> 00:03:39,450 be characterized by these three things. 80 00:03:39,450 --> 00:03:42,550 And then they go and they try to show you the three things. 81 00:03:42,550 --> 00:03:46,270 The figures, I think, are also very nice, 82 00:03:46,270 --> 00:03:48,010 in the sense that in many of the cases 83 00:03:48,010 --> 00:03:51,450 you could have shown the data just in the context of a table, 84 00:03:51,450 --> 00:03:53,850 where you said, oh, for each of these strains, this 85 00:03:53,850 --> 00:03:55,570 is the up regulation or so. 86 00:03:55,570 --> 00:03:57,570 But if you had done that, it would've been much, 87 00:03:57,570 --> 00:03:59,990 I think, less compelling, even though of course it's 88 00:03:59,990 --> 00:04:01,410 the same data. 89 00:04:01,410 --> 00:04:04,780 So I think this is a neat paper, in my opinion, 90 00:04:04,780 --> 00:04:07,370 both from the standpoint of the ideas that are being explored, 91 00:04:07,370 --> 00:04:09,527 but also because it highlights some of the things 92 00:04:09,527 --> 00:04:11,110 that you should be thinking about when 93 00:04:11,110 --> 00:04:12,450 you're writing your own papers. 94 00:04:12,450 --> 00:04:16,170 You want to try to make the ideas as clear as possible. 95 00:04:16,170 --> 00:04:18,050 You want to lay the groundwork so 96 00:04:18,050 --> 00:04:19,730 that what you're about to show is 97 00:04:19,730 --> 00:04:21,396 going to be-- the reader's going to feel 98 00:04:21,396 --> 00:04:23,190 is a really important thing. 99 00:04:23,190 --> 00:04:27,330 And then you want to take nice advantage of color 100 00:04:27,330 --> 00:04:29,290 and some legends that are there. 101 00:04:29,290 --> 00:04:33,280 So we'll kind of talk about all these issues as we go. 102 00:04:33,280 --> 00:04:37,280 But before we get started, can somebody-- 103 00:04:37,280 --> 00:04:40,470 there's a very real sense that in this field of decision 104 00:04:40,470 --> 00:04:47,000 making and systems biology that a lot of this research program 105 00:04:47,000 --> 00:04:49,830 is kind of driven by following classic ideas 106 00:04:49,830 --> 00:04:51,370 from other fields. 107 00:04:51,370 --> 00:04:54,940 And what would be the corresponding classic idea 108 00:04:54,940 --> 00:04:59,312 this paper is exploring? 109 00:04:59,312 --> 00:05:00,270 AUDIENCE: Conditioning. 110 00:05:00,270 --> 00:05:01,269 PROFESSOR: Conditioning. 111 00:05:01,269 --> 00:05:02,440 Right. 112 00:05:02,440 --> 00:05:05,062 And whose name do we associate with conditioning typically? 113 00:05:05,062 --> 00:05:05,770 AUDIENCE: Pavlov. 114 00:05:05,770 --> 00:05:06,420 PROFESSOR: Pavlov. 115 00:05:06,420 --> 00:05:06,920 All right. 116 00:05:06,920 --> 00:05:09,810 So I have not ever read these studies. 117 00:05:09,810 --> 00:05:11,660 Pavlovian conditioning. 118 00:05:14,229 --> 00:05:16,520 And I think that this is just-- it's good to highlight. 119 00:05:16,520 --> 00:05:18,061 This is something that you might have 120 00:05:18,061 --> 00:05:21,220 learned in your high school psychology class. 121 00:05:21,220 --> 00:05:26,500 And it's, again, a big idea, but it's not the kind of thing-- 122 00:05:26,500 --> 00:05:27,970 we've all heard-- well, we're going 123 00:05:27,970 --> 00:05:29,261 to talk about this in a moment. 124 00:05:29,261 --> 00:05:30,980 Many of us, I think, have heard of this. 125 00:05:30,980 --> 00:05:33,146 But this is an example of how you take something 126 00:05:33,146 --> 00:05:34,770 that you learned in high school and you 127 00:05:34,770 --> 00:05:39,475 make it useful to your daily life, or pseudo daily life. 128 00:05:39,475 --> 00:05:42,100 Because I think you'll see that there are many kind of examples 129 00:05:42,100 --> 00:05:44,280 of this throughout this literature, 130 00:05:44,280 --> 00:05:48,630 where someone takes an idea that is, in some ways, 131 00:05:48,630 --> 00:05:51,880 you open up a random textbook in introductory psychology, 132 00:05:51,880 --> 00:05:54,420 and you can just march through and try 133 00:05:54,420 --> 00:05:57,370 to see to what degree the ideas that were developed 134 00:05:57,370 --> 00:05:59,830 in the context of humans or animals, 135 00:05:59,830 --> 00:06:02,710 to what degree might they be relevant in the context of cell 136 00:06:02,710 --> 00:06:03,960 decision making? 137 00:06:03,960 --> 00:06:06,470 Can somebody just say what is this Pavlovian conditioning 138 00:06:06,470 --> 00:06:06,970 idea? 139 00:06:12,560 --> 00:06:15,860 Did you guys take high school psychology? 140 00:06:15,860 --> 00:06:17,190 I'm sure somebody did. 141 00:06:17,190 --> 00:06:17,790 No. 142 00:06:17,790 --> 00:06:18,290 All right. 143 00:06:18,290 --> 00:06:20,710 Well, incidentally, I recommend everybody 144 00:06:20,710 --> 00:06:22,760 should take a solid introductory psychology 145 00:06:22,760 --> 00:06:24,400 class if you have not done so. 146 00:06:24,400 --> 00:06:26,420 We offer a class 900. 147 00:06:26,420 --> 00:06:28,700 I'm sure that it's good and interesting. 148 00:06:28,700 --> 00:06:30,335 Yes. 149 00:06:30,335 --> 00:06:35,467 AUDIENCE: So there's two different events or stimuli, 150 00:06:35,467 --> 00:06:40,377 and in Pavlovian conditioning you give one and then 151 00:06:40,377 --> 00:06:43,330 always give the other one following. 152 00:06:43,330 --> 00:06:44,370 PROFESSOR: That's right. 153 00:06:44,370 --> 00:06:48,930 So there's some sense that what we might call-- 154 00:06:48,930 --> 00:06:52,080 there are two events of some sort. 155 00:06:52,080 --> 00:06:55,970 One is following the other one, so it's A and then followed 156 00:06:55,970 --> 00:07:00,150 by B. And people have done many different examples of this, 157 00:07:00,150 --> 00:07:03,680 but what's the one that we typically-- 158 00:07:03,680 --> 00:07:08,289 what's the classic experiment that Pavlov did? 159 00:07:08,289 --> 00:07:10,330 AUDIENCE: You ring a bell, and then you get food. 160 00:07:10,330 --> 00:07:10,650 PROFESSOR: Yeah. 161 00:07:10,650 --> 00:07:11,150 Right. 162 00:07:11,150 --> 00:07:20,860 So you ring a bell and then you give-- and this is dogs, 163 00:07:20,860 --> 00:07:23,940 at least in the version I remember. 164 00:07:23,940 --> 00:07:26,117 So you have some dogs, you do this thing 165 00:07:26,117 --> 00:07:28,450 where you ring the bell and then you give them the food. 166 00:07:28,450 --> 00:07:33,048 And what's the response that you're supposed to get? 167 00:07:33,048 --> 00:07:33,920 AUDIENCE: Drooling. 168 00:07:33,920 --> 00:07:34,878 PROFESSOR: Yeah, right. 169 00:07:34,878 --> 00:07:37,080 So the dog is supposed to start salivating. 170 00:07:37,080 --> 00:07:37,580 OK. 171 00:07:37,580 --> 00:07:39,460 So you might think your experiments 172 00:07:39,460 --> 00:07:41,190 are gross, but-- right. 173 00:07:41,190 --> 00:07:44,670 So ring a bell, and the idea is that you can train the dog 174 00:07:44,670 --> 00:07:47,365 to start salivating in response to the bell, rather than, 175 00:07:47,365 --> 00:07:48,990 of course, if you give the food they're 176 00:07:48,990 --> 00:07:50,270 going to start salivating. 177 00:07:50,270 --> 00:07:53,270 But here you can train the dog to start salivating 178 00:07:53,270 --> 00:07:55,340 in response to the food. 179 00:07:55,340 --> 00:07:59,055 So now, with training, this results in salivation. 180 00:08:01,835 --> 00:08:04,870 And did Pavlov show there was a fitness benefit associated 181 00:08:04,870 --> 00:08:07,430 with the salivation? 182 00:08:07,430 --> 00:08:10,480 Not that I'm not I'm aware of. 183 00:08:10,480 --> 00:08:13,110 But this is the classic story that we learned 184 00:08:13,110 --> 00:08:14,670 in introductory psychology. 185 00:08:14,670 --> 00:08:19,610 And the question is, can we apply a similar idea 186 00:08:19,610 --> 00:08:21,340 in the case of microbial decision making? 187 00:08:24,230 --> 00:08:29,330 Now was this-- can somebody say what 188 00:08:29,330 --> 00:08:33,510 they think was the author's contribution 189 00:08:33,510 --> 00:08:37,580 to the literature in the sense of-- via a vis this? 190 00:08:37,580 --> 00:08:41,450 How much of what-- well, we can be maybe more concrete. 191 00:08:41,450 --> 00:08:48,239 In the case of E. coli, what was known-- or maybe 192 00:08:48,239 --> 00:08:50,030 we should describe the experiment once more 193 00:08:50,030 --> 00:08:53,930 and then try to figure out what they did that was new new. 194 00:08:53,930 --> 00:08:56,714 At least can somebody kind of summarize the basic idea 195 00:08:56,714 --> 00:08:57,255 with E. coli? 196 00:08:59,950 --> 00:09:04,324 AUDIENCE: We know that they have a certain metabolic life cycle 197 00:09:04,324 --> 00:09:08,710 and that they exist in different environments. 198 00:09:08,710 --> 00:09:09,486 PROFESSOR: OK. 199 00:09:09,486 --> 00:09:11,610 So that E. coli existent in different environments. 200 00:09:11,610 --> 00:09:13,401 And what environments are you referring to? 201 00:09:13,401 --> 00:09:15,185 AUDIENCE: Gut and outside. 202 00:09:15,185 --> 00:09:16,310 PROFESSOR: Gut and outside. 203 00:09:16,310 --> 00:09:19,490 So if this is true-- then what is 204 00:09:19,490 --> 00:09:21,620 going to be your next statement based on this? 205 00:09:21,620 --> 00:09:25,103 AUDIENCE: Well, going through mammalian guts 206 00:09:25,103 --> 00:09:28,967 there's always a certain sequence of events. 207 00:09:28,967 --> 00:09:31,221 At the beginning, there's some acid, and then 208 00:09:31,221 --> 00:09:33,095 less acid and more food, and stuff like that. 209 00:09:35,730 --> 00:09:36,570 PROFESSOR: RIght. 210 00:09:36,570 --> 00:09:40,340 So in the context of E. coli, there's this idea 211 00:09:40,340 --> 00:09:43,160 that part of the life cycle of E. coli 212 00:09:43,160 --> 00:09:46,005 and some other of these microbes is 213 00:09:46,005 --> 00:09:49,850 to live inside the gut of mammals. 214 00:09:49,850 --> 00:09:51,625 And we're going to focus in particular 215 00:09:51,625 --> 00:09:55,490 on this aspect, the part of the life cycle that is in the gut. 216 00:09:55,490 --> 00:09:59,380 So let's just imagine, so E. coli in gut or entering gut, 217 00:09:59,380 --> 00:10:03,220 going through the gut, they might encounter 218 00:10:03,220 --> 00:10:06,700 some sort of typical environmental cues, 219 00:10:06,700 --> 00:10:10,740 or maybe orders of environmental exposure. 220 00:10:10,740 --> 00:10:13,680 And actually, even before this paper, 221 00:10:13,680 --> 00:10:20,530 there was another paper that was published by Saeed Tavazoie-- 222 00:10:20,530 --> 00:10:23,340 hard name to pronouce-- who was at Princeton at the time. 223 00:10:23,340 --> 00:10:25,950 So he had published a Science paper just the year 224 00:10:25,950 --> 00:10:29,620 before-- well, couple years before-- demonstrating 225 00:10:29,620 --> 00:10:30,565 a related phenomenon. 226 00:10:33,480 --> 00:10:35,880 What was it that this other paper had demonstrated? 227 00:10:35,880 --> 00:10:38,960 Does anybody remember from the introduction? 228 00:10:38,960 --> 00:10:41,090 So it wasn't the experiment that these guys did. 229 00:10:41,090 --> 00:10:43,650 But in the introduction they spent 230 00:10:43,650 --> 00:10:46,153 half of a paragraph describing this other experiment. 231 00:10:50,590 --> 00:10:53,280 Another kind of environmental cue 232 00:10:53,280 --> 00:10:56,150 that E. coli get when they enter the gut. 233 00:10:56,150 --> 00:10:58,352 So what's-- yeah. 234 00:10:58,352 --> 00:11:02,280 AUDIENCE: Isn't there a temperature increase and oxygen 235 00:11:02,280 --> 00:11:02,960 decrease? 236 00:11:02,960 --> 00:11:06,200 PROFESSOR: Temperature increase and oxygen decrease. 237 00:11:06,200 --> 00:11:14,420 So temperature goes up and oxygen goes down. 238 00:11:14,420 --> 00:11:16,190 And this makes sense. 239 00:11:16,190 --> 00:11:18,910 The inside of our bodies typically 240 00:11:18,910 --> 00:11:20,160 warmer than outside our body. 241 00:11:20,160 --> 00:11:22,500 Also, it's a relatively anaerobic environment 242 00:11:22,500 --> 00:11:25,510 inside the gut as compared to the ready availability 243 00:11:25,510 --> 00:11:28,180 of oxygen outside. 244 00:11:28,180 --> 00:11:31,630 So what is it that this other paper then showed? 245 00:11:47,399 --> 00:11:47,899 Yes. 246 00:11:47,899 --> 00:11:49,542 AUDIENCE: So I think they were trying 247 00:11:49,542 --> 00:11:51,843 to show a symmetric response [INAUDIBLE]. 248 00:11:55,294 --> 00:11:57,759 If adaptation in one situation occurred, 249 00:11:57,759 --> 00:12:02,689 then-- if it received one stimulus, 250 00:12:02,689 --> 00:12:04,425 it would adapt to both. 251 00:12:04,425 --> 00:12:05,050 PROFESSOR: Yes. 252 00:12:05,050 --> 00:12:05,550 OK. 253 00:12:05,550 --> 00:12:08,110 And the word adapt, we always have to be very careful, 254 00:12:08,110 --> 00:12:10,110 because it could mean two very different things. 255 00:12:10,110 --> 00:12:13,360 And in some cases, it might be that both-- possibly it 256 00:12:13,360 --> 00:12:15,750 might be true, but when you say that it receives 257 00:12:15,750 --> 00:12:20,910 one stimuluss-- so this we might call S1, 258 00:12:20,910 --> 00:12:24,820 and this is S2-- so these are stimuli that the cell 259 00:12:24,820 --> 00:12:28,356 senses-- are you saying that it might sense one, 260 00:12:28,356 --> 00:12:29,480 and then it adapts to both. 261 00:12:29,480 --> 00:12:31,800 When you say adapt, are you referring 262 00:12:31,800 --> 00:12:33,890 to evolutionary adaptation or something else? 263 00:12:33,890 --> 00:12:35,884 AUDIENCE: I'm referring to a single cell 264 00:12:35,884 --> 00:12:37,550 implementing the appropriate metabolic-- 265 00:12:37,550 --> 00:12:40,910 PROFESSOR: That sort of cell level adaptation. 266 00:12:40,910 --> 00:12:43,080 The sort of default thing is you might, say, well, 267 00:12:43,080 --> 00:12:47,240 if you get the stimuli one and increase the temperature, 268 00:12:47,240 --> 00:12:52,240 you should implement the program one or the response one that 269 00:12:52,240 --> 00:12:57,220 corresponds to some set of genes that are required 270 00:12:57,220 --> 00:13:00,670 to grow in high temperature. 271 00:13:00,670 --> 00:13:06,450 Now similarly with stimulus 2-- well, we'll say this is R2. 272 00:13:06,450 --> 00:13:10,460 So these could, in principal, have different gene network 273 00:13:10,460 --> 00:13:13,330 responses to these two environmental changes. 274 00:13:13,330 --> 00:13:16,280 But what they found is that actually, if you take E. coli 275 00:13:16,280 --> 00:13:19,240 and you expose it to increase in temperature, 276 00:13:19,240 --> 00:13:21,869 then not only does it activate the genes required 277 00:13:21,869 --> 00:13:23,785 to handle high temperature, but also activates 278 00:13:23,785 --> 00:13:29,902 the genes required handle low oxygen, and vice versa. 279 00:13:32,740 --> 00:13:35,290 Now there's a question, though. 280 00:13:35,290 --> 00:13:37,470 If you see this, it could just be the case 281 00:13:37,470 --> 00:13:40,240 that you've mislabeled what really is R1, R2. 282 00:13:40,240 --> 00:13:43,670 Maybe it's just a single R, single response 283 00:13:43,670 --> 00:13:45,910 that allows the population to adapt to both 284 00:13:45,910 --> 00:13:49,737 of those environmental changes. 285 00:13:49,737 --> 00:13:51,320 Does anybody remember what the authors 286 00:13:51,320 --> 00:13:54,640 did to show that that was maybe not what was going on? 287 00:14:10,247 --> 00:14:12,580 You can look at your sheet of paper while I--I just want 288 00:14:12,580 --> 00:14:16,740 to write down the reference in case anybody wants to look 289 00:14:16,740 --> 00:14:17,500 at this more. 290 00:14:17,500 --> 00:14:31,750 This is Tagkopoulos, Science 2008. 291 00:14:31,750 --> 00:14:33,590 What would you do if you wanted to show 292 00:14:33,590 --> 00:14:36,780 that this was a real regulatory strategy, 293 00:14:36,780 --> 00:14:40,960 this so-called symmetric anticipatory regulation? 294 00:15:04,040 --> 00:15:04,540 Yes. 295 00:15:04,540 --> 00:15:06,294 AUDIENCE: Would it be you could measure 296 00:15:06,294 --> 00:15:08,855 genes' pressure somehow, and then you 297 00:15:08,855 --> 00:15:11,662 could get the function of each 298 00:15:11,662 --> 00:15:12,370 PROFESSOR: Right. 299 00:15:12,370 --> 00:15:15,035 So you could expose to this stimulus, 300 00:15:15,035 --> 00:15:17,660 measure gene expression, expose to this stimulus, measures gene 301 00:15:17,660 --> 00:15:19,550 expression, and try to figure out, maybe, 302 00:15:19,550 --> 00:15:21,850 based on the difference which genes were more 303 00:15:21,850 --> 00:15:24,770 or less-- but the argument was somehow 304 00:15:24,770 --> 00:15:30,510 that in response to stimulus 1, there were some genes that 305 00:15:30,510 --> 00:15:32,059 were really unnecessary. 306 00:15:32,059 --> 00:15:33,600 So I guess you could go and you could 307 00:15:33,600 --> 00:15:36,930 do knockouts of each of these 100 genes, 308 00:15:36,930 --> 00:15:39,030 and then try to measure the effects. 309 00:15:39,030 --> 00:15:42,655 And that would be a nice direct-- 310 00:15:42,655 --> 00:15:45,030 and it's related to something that they did in this paper 311 00:15:45,030 --> 00:15:45,940 with the yeast. 312 00:15:45,940 --> 00:15:48,370 They went and looked at the gene expression profiles. 313 00:15:50,990 --> 00:15:55,230 What they did in this paper by Saeed 314 00:15:55,230 --> 00:15:58,350 was that they took the E. coli that 315 00:15:58,350 --> 00:16:01,490 displayed this, what they're arguing is a symmetric-- 316 00:16:01,490 --> 00:16:04,470 and based on some previous studies, I think, 317 00:16:04,470 --> 00:16:06,800 saying that, oh, this had been annotated to be good 318 00:16:06,800 --> 00:16:08,470 for temperature but not for oxygen and vice versa. 319 00:16:08,470 --> 00:16:10,680 So there were already hints that this kind of made sense. 320 00:16:10,680 --> 00:16:12,720 But what they did is they evolved the populations 321 00:16:12,720 --> 00:16:13,770 in just one of these environments 322 00:16:13,770 --> 00:16:15,478 and showed that they could decouple them. 323 00:16:18,710 --> 00:16:20,500 So that's at least making you feel 324 00:16:20,500 --> 00:16:25,740 that it's a little bit-- that maybe this was not 325 00:16:25,740 --> 00:16:28,692 due to just some sense that it's the same genes that 326 00:16:28,692 --> 00:16:31,150 are required for protection against these two environments, 327 00:16:31,150 --> 00:16:36,450 but rather that because they had evolved in this environment 328 00:16:36,450 --> 00:16:38,660 where, when they see an increase in temperature, 329 00:16:38,660 --> 00:16:42,324 they often see a decrease in oxygen. 330 00:16:42,324 --> 00:16:43,824 AUDIENCE: So when you say decoupled, 331 00:16:43,824 --> 00:16:48,050 you mean you evolve them in high-temperature environments? 332 00:16:48,050 --> 00:16:49,319 PROFESSOR: That's right. 333 00:16:49,319 --> 00:16:51,235 AUDIENCE: You expose them to decreased oxygen, 334 00:16:51,235 --> 00:16:52,050 and then they die? 335 00:16:52,050 --> 00:16:53,133 PROFESSOR: No, no, no, no. 336 00:16:53,133 --> 00:16:56,420 So just, they showed that if you evolve the population 337 00:16:56,420 --> 00:17:00,370 so that it receives this stimulus without this stimulus, 338 00:17:00,370 --> 00:17:02,850 then it removes this arrow, that it 339 00:17:02,850 --> 00:17:05,500 stops activating these genes. 340 00:17:05,500 --> 00:17:07,500 And then I think they also did look 341 00:17:07,500 --> 00:17:08,815 at the functional consequences. 342 00:17:08,815 --> 00:17:10,190 But the first order thing is just 343 00:17:10,190 --> 00:17:12,079 that they saw that they could remove 344 00:17:12,079 --> 00:17:16,140 one of these arrows, which is related to something that they 345 00:17:16,140 --> 00:17:18,930 do here as well. 346 00:17:18,930 --> 00:17:22,810 Now I think that it's worth thinking 347 00:17:22,810 --> 00:17:27,690 about the question of whether this study, published 348 00:17:27,690 --> 00:17:30,980 in Science in 2008, do you think that it's 349 00:17:30,980 --> 00:17:34,530 a complete scoop of the paper that you just read, 350 00:17:34,530 --> 00:17:43,880 which is Amir Mitchell's paper published in Nature in 2009? 351 00:17:43,880 --> 00:17:44,570 No. 352 00:17:44,570 --> 00:17:45,070 OK. 353 00:17:45,070 --> 00:17:46,819 And on one level you say, well, they still 354 00:17:46,819 --> 00:17:51,250 got it published in Nature, so therefore it was not a scoop. 355 00:17:51,250 --> 00:17:58,260 But I can tell you that Amir was very worried, 356 00:17:58,260 --> 00:18:02,260 because he had been working on this paper for many years. 357 00:18:02,260 --> 00:18:05,020 And they really studied both E. coli and yeasts. 358 00:18:05,020 --> 00:18:06,590 It was a major thing. 359 00:18:06,590 --> 00:18:08,850 And then this paper came out, and I'm 360 00:18:08,850 --> 00:18:11,260 sure that he was despondent. 361 00:18:11,260 --> 00:18:12,874 And there will be times where you 362 00:18:12,874 --> 00:18:14,540 feel despondent as a result of something 363 00:18:14,540 --> 00:18:17,990 like this coming out, because it really is a very similar idea. 364 00:18:17,990 --> 00:18:20,850 The E. coli stuff is very similar, 365 00:18:20,850 --> 00:18:27,200 because it's same organism-- the idea is, in the gut. 366 00:18:27,200 --> 00:18:30,024 So in what ways do you think that their paper is different, 367 00:18:30,024 --> 00:18:31,440 and also, what strategies did they 368 00:18:31,440 --> 00:18:34,370 follow in order to differentiate their paper? 369 00:18:37,540 --> 00:18:38,296 Yeah. 370 00:18:38,296 --> 00:18:40,810 AUDIENCE: Maybe potentially trying 371 00:18:40,810 --> 00:18:46,205 to find differing regulation mechanisms, so they said, well, 372 00:18:46,205 --> 00:18:48,065 there's a symmetric regulation mechanism 373 00:18:48,065 --> 00:18:49,565 that someone talked about last year, 374 00:18:49,565 --> 00:18:51,990 and we've got the asymmetric one. 375 00:18:51,990 --> 00:18:52,990 PROFESSOR: That's right. 376 00:18:52,990 --> 00:18:56,620 And I think that this is-- of course, 377 00:18:56,620 --> 00:18:58,860 we cannot replay history. 378 00:18:58,860 --> 00:19:03,410 But I think that they do a very nice thing in figure one 379 00:19:03,410 --> 00:19:07,370 of laying the groundwork to make it very clear 380 00:19:07,370 --> 00:19:10,150 that their work is different from Saeed's work. 381 00:19:10,150 --> 00:19:13,740 And the thing is that if this had not been published, 382 00:19:13,740 --> 00:19:16,840 they maybe would have framed their work differently. 383 00:19:16,840 --> 00:19:18,870 They might not have had this difference 384 00:19:18,870 --> 00:19:20,940 between a symmetrical anticipatory regulation 385 00:19:20,940 --> 00:19:23,310 and the asymmetrical. 386 00:19:23,310 --> 00:19:25,850 But by laying things out that way, 387 00:19:25,850 --> 00:19:28,200 it clarifies that, oh yeah, these are different things. 388 00:19:28,200 --> 00:19:30,580 And they're going to explore this one, 389 00:19:30,580 --> 00:19:33,240 whereas Saeed's group had explored 390 00:19:33,240 --> 00:19:34,672 a rather different mechanism. 391 00:19:34,672 --> 00:19:36,880 But then, of course, depending on how you look at it, 392 00:19:36,880 --> 00:19:38,379 they're either more or less similar. 393 00:19:38,379 --> 00:19:41,140 But I think that they do a very nice job of kind of right 394 00:19:41,140 --> 00:19:43,154 at the beginning setting out how their work is 395 00:19:43,154 --> 00:19:44,070 going to be different. 396 00:19:50,170 --> 00:19:52,720 So they explained that there's the simple direct sensing 397 00:19:52,720 --> 00:19:54,720 mechanisms. 398 00:19:54,720 --> 00:19:59,370 You'd say that you just have different stimuli that 399 00:19:59,370 --> 00:20:01,080 lead to different responses. 400 00:20:01,080 --> 00:20:02,354 That's fine. 401 00:20:02,354 --> 00:20:04,270 They also have the stochastic switching, which 402 00:20:04,270 --> 00:20:05,561 we're going to talk about more. 403 00:20:10,620 --> 00:20:12,540 The idea here is that even though we 404 00:20:12,540 --> 00:20:16,582 might be getting different stimuli at some rates, 405 00:20:16,582 --> 00:20:18,790 we would still switch into these different responses. 406 00:20:22,240 --> 00:20:25,000 And we'll talk about various reasons that this might happen. 407 00:20:25,000 --> 00:20:26,541 Incidentally, these things don't have 408 00:20:26,541 --> 00:20:28,070 to happen at the same rates. 409 00:20:28,070 --> 00:20:31,570 So it could be that stimulus one it primarily does R1, 410 00:20:31,570 --> 00:20:35,250 whereas stimulus 2 primarily does R2, for example. 411 00:20:35,250 --> 00:20:37,985 If you'd like, these arrows could have different strengths. 412 00:20:40,800 --> 00:20:43,380 But then what they do is they point out that there really 413 00:20:43,380 --> 00:20:48,572 are-- well, this symmetric anticipatory regulation, 414 00:20:48,572 --> 00:20:50,030 where it's crossed, maybe I'll just 415 00:20:50,030 --> 00:20:51,238 write out the asymmetric one. 416 00:21:04,960 --> 00:21:06,670 Can somebody say why it is that we 417 00:21:06,670 --> 00:21:09,350 might want to do the asymmetrical as compared 418 00:21:09,350 --> 00:21:10,645 to the symmetrical? 419 00:21:30,425 --> 00:21:32,900 Yes. 420 00:21:32,900 --> 00:21:36,860 AUDIENCE: If both of the stimuli are appearing at the same time, 421 00:21:36,860 --> 00:21:39,335 then you might want to have a symmetric anticipatory 422 00:21:39,335 --> 00:21:40,322 regulation? 423 00:21:40,322 --> 00:21:41,030 PROFESSOR: Great. 424 00:21:41,030 --> 00:21:42,770 So the important thin is here that this 425 00:21:42,770 --> 00:21:47,055 is, because it's at the same time, the two stimuli kind 426 00:21:47,055 --> 00:21:53,230 of come together, or very similar times, whereas this, 427 00:21:53,230 --> 00:21:55,330 if there's a clear temporal order. 428 00:21:59,020 --> 00:22:01,030 But you could think about this in some other way 429 00:22:01,030 --> 00:22:02,090 as well, maybe. 430 00:22:02,090 --> 00:22:02,850 Can somebody-- 431 00:22:10,261 --> 00:22:11,760 AUDIENCE: It's along the same lines, 432 00:22:11,760 --> 00:22:17,070 but there's a cost to having S2 also performing. 433 00:22:17,070 --> 00:22:18,195 Then if you ever use that-- 434 00:22:21,162 --> 00:22:21,870 PROFESSOR: Right. 435 00:22:21,870 --> 00:22:23,450 So depending on the costs, it could 436 00:22:23,450 --> 00:22:28,540 change what's sort of optimal, perhaps, right? 437 00:22:28,540 --> 00:22:30,700 And I think it's also perhaps worth just saying 438 00:22:30,700 --> 00:22:32,520 that it may also depend upon how frequent 439 00:22:32,520 --> 00:22:34,520 these different environmental perturbations are, 440 00:22:34,520 --> 00:22:36,630 because you can also imagine situations 441 00:22:36,630 --> 00:22:43,470 where it's very common that the temperature goes up, 442 00:22:43,470 --> 00:22:46,180 but it's very rare that oxygen goes down. 443 00:22:46,180 --> 00:22:50,431 But there could be a sense in which this is, basically, 444 00:22:50,431 --> 00:22:51,930 based on the logic I just said, this 445 00:22:51,930 --> 00:22:53,806 would be then less predictive of this 446 00:22:53,806 --> 00:22:54,930 than this would be of this. 447 00:22:54,930 --> 00:22:56,450 Do you guys see what I'm saying? 448 00:22:56,450 --> 00:22:58,908 Because it could be the case that any time that oxygen goes 449 00:22:58,908 --> 00:23:01,810 down, then you really know that that means that you're entering 450 00:23:01,810 --> 00:23:04,740 somebody's gut and you also better activate this, 451 00:23:04,740 --> 00:23:07,340 whereas it could also be the case that if you sit out 452 00:23:07,340 --> 00:23:09,190 in the sun, you get hot, it doesn't mean 453 00:23:09,190 --> 00:23:10,481 that oxygen's going to go down. 454 00:23:10,481 --> 00:23:13,750 So it could be that there is more or less information, 455 00:23:13,750 --> 00:23:16,380 or that each of the stimuli have more or less information 456 00:23:16,380 --> 00:23:17,590 about another stimuli. 457 00:23:17,590 --> 00:23:19,950 So even if there's no temporal order, 458 00:23:19,950 --> 00:23:22,900 it could be the case that something like that 459 00:23:22,900 --> 00:23:26,010 could lead to a situation where this might be optimal. 460 00:23:29,164 --> 00:23:33,625 AUDIENCE: So it seems people try to make 461 00:23:33,625 --> 00:23:37,485 formal statements about information in this context, 462 00:23:37,485 --> 00:23:40,275 just like they did in the context of stochastic theory. 463 00:23:43,540 --> 00:23:45,990 PROFESSOR: So there's some of it. 464 00:23:45,990 --> 00:23:48,280 I haven't seen anything that's quite as 465 00:23:48,280 --> 00:23:51,740 clear and formal as what Edo Kussell did in that Science 466 00:23:51,740 --> 00:23:53,070 paper you guys just read. 467 00:23:53,070 --> 00:23:56,250 But for example, Amir did do a paper, 468 00:23:56,250 --> 00:24:00,600 he wrote a PNAS paper basically laying out this model that's 469 00:24:00,600 --> 00:24:03,520 sort of sketched at the end of this paper a little bit 470 00:24:03,520 --> 00:24:05,830 in more detail in terms of explaining, 471 00:24:05,830 --> 00:24:08,310 depending on the probability that something occurs 472 00:24:08,310 --> 00:24:10,020 and the time, so kind of exploring 473 00:24:10,020 --> 00:24:11,680 this a bit more, but not as formal 474 00:24:11,680 --> 00:24:14,645 in terms of information as-- 475 00:24:14,645 --> 00:24:16,270 AUDIENCE: So it's not quite clear to me 476 00:24:16,270 --> 00:24:20,680 why you would have a symmetric anticipatory regulation 477 00:24:20,680 --> 00:24:23,130 comparing direct sensing. 478 00:24:23,130 --> 00:24:28,070 If you have direct sensing, you can do the same thing. 479 00:24:28,070 --> 00:24:29,610 PROFESSOR: That's right. 480 00:24:29,610 --> 00:24:30,360 I think it's true. 481 00:24:30,360 --> 00:24:33,730 But it could also be that because these things are 482 00:24:33,730 --> 00:24:37,940 both fluctuating and they are measured 483 00:24:37,940 --> 00:24:41,922 with some noise and the environment, 484 00:24:41,922 --> 00:24:44,130 there could be a sense that-- because you can imagine 485 00:24:44,130 --> 00:24:46,880 a world in which these two things are 486 00:24:46,880 --> 00:24:50,210 both uncertain predictors of whether you're in the gut. 487 00:24:50,210 --> 00:24:52,690 But if they both go, then that really 488 00:24:52,690 --> 00:24:58,620 means something, in which case then it would make sense 489 00:24:58,620 --> 00:25:03,120 to really look at both of them and activate both more strongly 490 00:25:03,120 --> 00:25:05,390 as a result of seeing both. 491 00:25:05,390 --> 00:25:07,186 AUDIENCE: But it could also make you 492 00:25:07,186 --> 00:25:09,800 more susceptible to trigger a response that you 493 00:25:09,800 --> 00:25:11,760 don't necessarily want. 494 00:25:11,760 --> 00:25:12,890 PROFESSOR: That's true. 495 00:25:12,890 --> 00:25:16,420 But if you say, oh, well this happens at a low rate 496 00:25:16,420 --> 00:25:17,920 and this happens at a low rate, then 497 00:25:17,920 --> 00:25:20,720 the probability that both of them happen at the same time 498 00:25:20,720 --> 00:25:21,850 would be very small. 499 00:25:21,850 --> 00:25:24,120 So your rate of false positive, in that sense, 500 00:25:24,120 --> 00:25:27,760 would be-- you can really reduce the rate of false positives 501 00:25:27,760 --> 00:25:29,790 by integrating both pieces of information. 502 00:25:38,380 --> 00:25:40,980 And this paper, just to remind ourselves, 503 00:25:40,980 --> 00:25:42,980 this paper is primarily exploring 504 00:25:42,980 --> 00:25:46,280 the idea of asymmetrical anticipatory regulation. 505 00:25:46,280 --> 00:25:48,340 And their argument is because there's 506 00:25:48,340 --> 00:25:54,420 a typical temporal order to how E. Coli that are ingested 507 00:25:54,420 --> 00:25:57,640 encounter different sugar. 508 00:25:57,640 --> 00:26:02,560 And the claim in the paper is that bacteria will typically 509 00:26:02,560 --> 00:26:04,060 see what before what? 510 00:26:11,365 --> 00:26:12,826 AUDIENCE: Lactose before maltose. 511 00:26:12,826 --> 00:26:13,617 PROFESSOR: Correct. 512 00:26:13,617 --> 00:26:18,660 So the claim is that in the gut, the bacteria will see lactose 513 00:26:18,660 --> 00:26:22,170 before they see maltose. 514 00:26:22,170 --> 00:26:25,880 Now in order to evolve this asymmetrical anticipatory 515 00:26:25,880 --> 00:26:30,812 regulation, does it have to be that you kind of always 516 00:26:30,812 --> 00:26:32,270 see lactose before you see maltose? 517 00:26:39,730 --> 00:26:43,450 So let's imagine that only 10% of the time that the bacteria 518 00:26:43,450 --> 00:26:46,550 see lactose-- or sorry, only 10% of the time 519 00:26:46,550 --> 00:26:49,540 that the bacteria encounter maltose that they actually saw 520 00:26:49,540 --> 00:26:50,420 lactose before. 521 00:26:53,420 --> 00:26:56,010 Since 10% is much less than 1, should that totally 522 00:26:56,010 --> 00:26:59,070 scotch this whole mechanism? 523 00:26:59,070 --> 00:26:59,570 No. 524 00:26:59,570 --> 00:27:01,880 And can you say why not? 525 00:27:01,880 --> 00:27:04,750 AUDIENCE: Every time they see lactose they see maltose. 526 00:27:04,750 --> 00:27:05,750 PROFESSOR: That's right. 527 00:27:05,750 --> 00:27:07,910 In some ways, that's not the right way 528 00:27:07,910 --> 00:27:12,840 to think about it, but in terms of a cost-benefit type 529 00:27:12,840 --> 00:27:16,400 argument, why is it that maybe it's OK in principle 530 00:27:16,400 --> 00:27:19,220 that maltose is frequently observed without lactose? 531 00:27:25,891 --> 00:27:28,326 AUDIENCE: There's no difference between the two strains 532 00:27:28,326 --> 00:27:30,032 if you see maltose that way. 533 00:27:30,032 --> 00:27:30,740 PROFESSOR: Right. 534 00:27:30,740 --> 00:27:33,440 And I think this is a kind of an interesting, important point, 535 00:27:33,440 --> 00:27:37,180 which is that if you imagine a competition between the strain 536 00:27:37,180 --> 00:27:41,360 that just does, we can imagine, direct sensing versus we'll 537 00:27:41,360 --> 00:27:42,440 call that AAR. 538 00:27:42,440 --> 00:27:49,090 So direct sensing versus asymmetrical anticipatory 539 00:27:49,090 --> 00:27:50,180 regulation. 540 00:27:50,180 --> 00:27:56,110 Now if you imagine the situation where you encounter 541 00:27:56,110 --> 00:27:59,350 maltose without encountering lactose, then these two 542 00:27:59,350 --> 00:28:02,516 strategies behave essentially identically, 543 00:28:02,516 --> 00:28:04,140 because they both just see the maltose, 544 00:28:04,140 --> 00:28:05,650 they're both surprised. 545 00:28:05,650 --> 00:28:08,080 They both activate the mall genes, everything's fine. 546 00:28:10,800 --> 00:28:12,540 However, it's in this other direction 547 00:28:12,540 --> 00:28:13,873 that you have to think about it. 548 00:28:13,873 --> 00:28:16,420 So the real question is, when you encounter lactose, 549 00:28:16,420 --> 00:28:19,230 you want to think about, well, what fraction of the time 550 00:28:19,230 --> 00:28:23,220 and how long a delay is there before you encounter maltose? 551 00:28:23,220 --> 00:28:28,270 Because if it's the case that you encounter lactose a lot, 552 00:28:28,270 --> 00:28:32,140 and only infrequently do you then come on to maltose, 553 00:28:32,140 --> 00:28:34,090 that's going to be a problem for evolving this 554 00:28:34,090 --> 00:28:36,190 into asymmetrical anticipatory regulation, 555 00:28:36,190 --> 00:28:38,900 or it would not necessarily be optimal to do so, 556 00:28:38,900 --> 00:28:41,190 because in that case you're going to be activating 557 00:28:41,190 --> 00:28:45,150 all those mall genes in advance, and frequently it 558 00:28:45,150 --> 00:28:45,920 won't be useful. 559 00:28:49,690 --> 00:28:54,730 And this is relevant, perhaps, just because what fraction 560 00:28:54,730 --> 00:28:58,112 of the time do you think that E. coli going to the gut 561 00:28:58,112 --> 00:28:59,070 will encounter lactose? 562 00:29:11,460 --> 00:29:13,840 We should be able to make a reasonable estimate of this, 563 00:29:13,840 --> 00:29:14,620 is my claim. 564 00:29:22,330 --> 00:29:22,830 All right. 565 00:29:22,830 --> 00:29:25,360 Where does lactose come from? 566 00:29:25,360 --> 00:29:28,160 Lactose comes from milk. 567 00:29:36,720 --> 00:29:37,800 Which animals drink milk? 568 00:29:40,660 --> 00:29:41,690 Well, right, OK. 569 00:29:41,690 --> 00:29:43,110 So we're thinking about mammals. 570 00:29:43,110 --> 00:29:46,142 Just take a random mammal of your choice. 571 00:29:46,142 --> 00:29:47,600 You might want to not choose humans 572 00:29:47,600 --> 00:29:50,020 just because we're a little bit atypical in this regard, 573 00:29:50,020 --> 00:29:51,970 although even there I think the numbers 574 00:29:51,970 --> 00:29:53,220 end up being kind of the same. 575 00:29:53,220 --> 00:30:03,680 But imagine a dog or a bear or your favorite mammal. 576 00:30:03,680 --> 00:30:06,890 Which of them are drinking milk? 577 00:30:06,890 --> 00:30:07,740 The young, right? 578 00:30:07,740 --> 00:30:08,239 OK. 579 00:30:08,239 --> 00:30:12,190 So there's some period of one's life where you drink milk, 580 00:30:12,190 --> 00:30:15,000 and then as an adult you typically don't. 581 00:30:15,000 --> 00:30:17,290 I guess all I'm saying is that if you 582 00:30:17,290 --> 00:30:20,960 imagine the bacteria entering into a mammalian gut, 583 00:30:20,960 --> 00:30:23,350 it might be 20%. 584 00:30:23,350 --> 00:30:26,560 I don't know what, but it's basically 585 00:30:26,560 --> 00:30:27,730 proportional to the years. 586 00:30:27,730 --> 00:30:31,080 The kids, they may eat more dirty things. 587 00:30:31,080 --> 00:30:32,890 So you have to wait or whatever. 588 00:30:32,890 --> 00:30:35,970 The point is that they're going to encounter lactose 589 00:30:35,970 --> 00:30:41,650 if they enter the gut of a baby or young child, 590 00:30:41,650 --> 00:30:45,260 whereas if you enter an adult, then you won't see lactose, 591 00:30:45,260 --> 00:30:48,650 and you'll just maybe then directly see the maltose. 592 00:30:48,650 --> 00:30:50,670 But there's no real cost associated 593 00:30:50,670 --> 00:30:52,910 with missing that signal vis a vis 594 00:30:52,910 --> 00:30:54,160 the so-called direct strategy. 595 00:31:02,900 --> 00:31:06,715 I think that they do lay out some other kind of conditions 596 00:31:06,715 --> 00:31:10,560 that they think would be reasonable to demonstrate 597 00:31:10,560 --> 00:31:15,830 if you want to do the asymmetrical anticipatory 598 00:31:15,830 --> 00:31:17,370 regulation. 599 00:31:17,370 --> 00:31:22,010 And once again, I think that this is both very clear 600 00:31:22,010 --> 00:31:23,440 and also just lays the groundwork 601 00:31:23,440 --> 00:31:26,310 for the reader in the sense that they tell you 602 00:31:26,310 --> 00:31:29,460 what they would need to show in order for you to be convinced. 603 00:31:29,460 --> 00:31:31,960 And then they go and they do those things. 604 00:31:31,960 --> 00:31:34,330 Now in principle, you could argue 605 00:31:34,330 --> 00:31:35,700 about one or the other things. 606 00:31:35,700 --> 00:31:37,570 But I think they do such a nice job 607 00:31:37,570 --> 00:31:40,400 of being clear about what would be needed that you don't even 608 00:31:40,400 --> 00:31:42,291 think to object about any of it. 609 00:31:42,291 --> 00:31:44,790 And of course they're not going to bring up the three things 610 00:31:44,790 --> 00:31:45,940 that they would need to demonstrate 611 00:31:45,940 --> 00:31:48,315 unless they actually were able to demonstrate those three 612 00:31:48,315 --> 00:31:49,470 things. 613 00:31:49,470 --> 00:31:51,720 So the three things that they claim 614 00:31:51,720 --> 00:31:58,810 is that for this to have been an evolved trait that pre-exposure 615 00:31:58,810 --> 00:32:08,180 to S1 increases the fitness under S2 or R2-- 616 00:32:08,180 --> 00:32:11,660 I guess in this case-- that's right. 617 00:32:11,660 --> 00:32:14,840 So getting S1 helps you in S2. 618 00:32:14,840 --> 00:32:17,504 So seeing lactose helps you in maltose. 619 00:32:17,504 --> 00:32:18,920 Indeed they're going to show that. 620 00:32:21,940 --> 00:32:25,780 Second, that there's some cost associated 621 00:32:25,780 --> 00:32:29,170 with up regulating that, in this case 622 00:32:29,170 --> 00:32:31,990 up regulating the mall genes. 623 00:32:31,990 --> 00:32:35,600 And third, there's this idea of specificity. 624 00:32:35,600 --> 00:32:39,110 Now the claim is that it's really 625 00:32:39,110 --> 00:32:41,710 S1 that is activating this R2 response, 626 00:32:41,710 --> 00:32:45,010 and that various other S's, in this case other sugar sources, 627 00:32:45,010 --> 00:32:46,790 maybe should not pre-activate. 628 00:32:56,260 --> 00:32:59,120 Now in figure 1, they basically are 629 00:32:59,120 --> 00:33:05,950 exploring the difference between this arrow and this one. 630 00:33:05,950 --> 00:33:08,710 So if you expose the cells to maltose 631 00:33:08,710 --> 00:33:12,370 you activate the mall genes at a high level. 632 00:33:12,370 --> 00:33:15,030 Is it required that exposure to lactose up 633 00:33:15,030 --> 00:33:20,570 regulates the mall genes to the same level as maltose? 634 00:33:20,570 --> 00:33:21,440 No. 635 00:33:21,440 --> 00:33:23,273 Was it at the same level as maltose? 636 00:33:23,273 --> 00:33:23,772 No. 637 00:33:26,570 --> 00:33:29,750 But the idea is it up regulates it some, as compared 638 00:33:29,750 --> 00:33:31,340 to other carbon sources. 639 00:33:37,760 --> 00:33:50,520 Now they also made a comment about whether this arrow, 640 00:33:50,520 --> 00:33:53,010 the so-called anticipatory regulation, 641 00:33:53,010 --> 00:33:55,010 whether it was specific to their E. coli strain, 642 00:33:55,010 --> 00:33:57,195 and their claim was that it was or was not? 643 00:34:03,349 --> 00:34:05,640 Would it be good for it to be specific to their strain, 644 00:34:05,640 --> 00:34:06,140 or bad? 645 00:34:08,620 --> 00:34:12,580 Probably bad, although the fact that-- but then they 646 00:34:12,580 --> 00:34:16,290 say, oh, it's not specific to our strain. 647 00:34:16,290 --> 00:34:20,150 And then they cite another paper, which-- and this 648 00:34:20,150 --> 00:34:21,650 is, again, a nice way of phrasing 649 00:34:21,650 --> 00:34:25,082 it because they're adding strength to their paper 650 00:34:25,082 --> 00:34:27,290 because they're saying, oh, it's not just our strain. 651 00:34:27,290 --> 00:34:29,030 It's in other strains as well. 652 00:34:29,030 --> 00:34:31,567 So you read that sentence and say yes, that's good. 653 00:34:31,567 --> 00:34:33,150 Of course you could read that sentence 654 00:34:33,150 --> 00:34:36,290 and say, oh, but somebody else already showed this? 655 00:34:36,290 --> 00:34:40,440 But this is, again-- and I don't have 656 00:34:40,440 --> 00:34:43,650 any real complaint about that, because I think what they've 657 00:34:43,650 --> 00:34:47,020 done is they've really brought a set of measurements 658 00:34:47,020 --> 00:34:49,110 together in one conceptual framework 659 00:34:49,110 --> 00:34:52,260 that I think is rather compelling. 660 00:34:52,260 --> 00:34:53,969 Now of course, you can always argue about 661 00:34:53,969 --> 00:34:55,780 whether each of these examples really 662 00:34:55,780 --> 00:34:58,460 did evolve for this purpose or not and so forth. 663 00:34:58,460 --> 00:35:00,190 But just because somebody else had 664 00:35:00,190 --> 00:35:03,810 seen that there was some cross-regulation of the mall 665 00:35:03,810 --> 00:35:05,091 genes in response to lactose. 666 00:35:05,091 --> 00:35:06,840 Doesn't mean that this paper is worthless. 667 00:35:11,860 --> 00:35:14,460 But it is interesting to note that somebody else had already 668 00:35:14,460 --> 00:35:14,960 seen this. 669 00:35:18,110 --> 00:35:21,610 Moreover, in figure one they compare 670 00:35:21,610 --> 00:35:27,310 this asymmetrical anticipatory regulation for their wild type 671 00:35:27,310 --> 00:35:29,182 strain as compared to another strain. 672 00:35:29,182 --> 00:35:30,640 And what other strain did they use? 673 00:35:36,630 --> 00:35:38,580 Yes. 674 00:35:38,580 --> 00:35:43,680 AUDIENCE: You didn't expose the cells to maltose 675 00:35:43,680 --> 00:35:44,962 after the lactose [INAUDIBLE]. 676 00:35:44,962 --> 00:35:45,670 PROFESSOR: Right. 677 00:35:45,670 --> 00:35:51,550 They evolved the cells in lactose for 500 generations. 678 00:35:54,345 --> 00:35:55,470 And then what did they see? 679 00:36:00,915 --> 00:36:01,910 AUDIENCE: The arrow. 680 00:36:01,910 --> 00:36:02,910 PROFESSOR: That's right. 681 00:36:02,910 --> 00:36:06,890 When they saw was that evolution-- and this 682 00:36:06,890 --> 00:36:13,110 is in lactose-- removed that. 683 00:36:16,910 --> 00:36:19,485 Now 500 generations, how long do you think 684 00:36:19,485 --> 00:36:20,610 this is going to take them? 685 00:36:27,820 --> 00:36:28,900 AUDIENCE: A few months. 686 00:36:28,900 --> 00:36:29,900 PROFESSOR: A few months. 687 00:36:33,090 --> 00:36:34,900 Have you guys ever heard of any other paper 688 00:36:34,900 --> 00:36:43,540 where they evolved E. coli in lactose for 500 generations 689 00:36:43,540 --> 00:36:45,590 in your extensive reading of the literature 690 00:36:45,590 --> 00:36:47,425 that you've engaged in over this semester? 691 00:36:53,231 --> 00:36:54,710 AUDIENCE: The cost-benefit paper. 692 00:36:54,710 --> 00:36:57,450 PROFESSOR: The cost-benefit paper. 693 00:36:57,450 --> 00:37:02,230 So it's fascinating to note that this sentence, they said, 694 00:37:02,230 --> 00:37:04,520 "We have examined laboratory-evolved strains 695 00:37:04,520 --> 00:37:06,180 of E. coli. 696 00:37:06,180 --> 00:37:08,780 Did they say that they did the evolution? 697 00:37:08,780 --> 00:37:10,680 No, they didn't. 698 00:37:10,680 --> 00:37:16,440 And indeed, if you follow Citation 14 over here, 699 00:37:16,440 --> 00:37:19,760 you can say, oh, right. 700 00:37:19,760 --> 00:37:24,790 So it's another paper by Uri and company. 701 00:37:24,790 --> 00:37:27,430 So I don't know if it was the exact same experiment, 702 00:37:27,430 --> 00:37:29,940 but the idea is that-- so I think 703 00:37:29,940 --> 00:37:32,070 what they did is, they took the strains that 704 00:37:32,070 --> 00:37:34,440 were evolved for another purpose, and that's fine. 705 00:37:34,440 --> 00:37:37,790 If somebody else already did the daily dilutions 706 00:37:37,790 --> 00:37:40,330 for three months that is the exact same that you were going 707 00:37:40,330 --> 00:37:42,640 to do, then you might as well just 708 00:37:42,640 --> 00:37:44,310 analyze their evolved strains, right? 709 00:37:44,310 --> 00:37:47,530 There's nothing that says you have to go and repeat the work. 710 00:37:47,530 --> 00:37:50,150 You would get the same thing, maybe, presumably. 711 00:37:50,150 --> 00:37:51,750 So this is just worth highlighting. 712 00:37:51,750 --> 00:37:53,330 But it makes the paper much stronger 713 00:37:53,330 --> 00:37:54,960 that they were able to do that. 714 00:37:54,960 --> 00:37:57,760 So it's good to keep in mind that in some cases, what you 715 00:37:57,760 --> 00:38:01,120 need to do to make your point has already been done 716 00:38:01,120 --> 00:38:02,920 or has already been evolved and so forth. 717 00:38:02,920 --> 00:38:04,560 So now what we have is a situation 718 00:38:04,560 --> 00:38:07,950 where somebody else, maybe, has already seen 719 00:38:07,950 --> 00:38:11,610 that lactose activates maltose. 720 00:38:11,610 --> 00:38:14,730 Somebody else already did this evolution on lactose. 721 00:38:14,730 --> 00:38:18,340 They didn't yet show that this regulation was removed. 722 00:38:18,340 --> 00:38:20,340 So that was something that these guys had to do, 723 00:38:20,340 --> 00:38:23,150 and indeed, it looks like it's removed. 724 00:38:23,150 --> 00:38:27,820 But once again, I think there's a sense that they're really 725 00:38:27,820 --> 00:38:30,570 looking at this problem from kind of a new, maybe 726 00:38:30,570 --> 00:38:32,219 comprehensive way that allows them 727 00:38:32,219 --> 00:38:33,635 to see the connections between all 728 00:38:33,635 --> 00:38:34,801 these different experiments. 729 00:38:34,801 --> 00:38:36,830 And then they are doing new measurements. 730 00:38:36,830 --> 00:38:39,590 But the measurements themselves are perhaps relatively simple, 731 00:38:39,590 --> 00:38:41,160 but the trick is knowing what measurements 732 00:38:41,160 --> 00:38:42,690 you need to do in order to say something 733 00:38:42,690 --> 00:38:43,689 that's very interesting. 734 00:38:52,610 --> 00:38:55,060 So in figure three, what they went and they did is 735 00:38:55,060 --> 00:38:57,310 they asked about this question of whether there really 736 00:38:57,310 --> 00:39:01,260 is a fitness benefit associated with it. 737 00:39:01,260 --> 00:39:04,120 Now I think this is another case where 738 00:39:04,120 --> 00:39:07,450 you could have made this figure in the form of a table, 739 00:39:07,450 --> 00:39:11,074 and it just would not be very compelling. 740 00:39:11,074 --> 00:39:13,740 So you'd look at it and you just wouldn't be very excited by it, 741 00:39:13,740 --> 00:39:17,440 whereas in this case, they have the figure, which 742 00:39:17,440 --> 00:39:20,550 are these fitness advantages associated with whether you 743 00:39:20,550 --> 00:39:23,290 take the strains, you first expose them to lactose and then 744 00:39:23,290 --> 00:39:25,360 go to maltose, or you go from maltose 745 00:39:25,360 --> 00:39:29,810 to lactose, or in this case GAL to maltose, sucrose to maltose. 746 00:39:29,810 --> 00:39:34,500 What they showed was it was a selective benefit there, 747 00:39:34,500 --> 00:39:38,870 that the cells really only gain the benefit of pre-exposure 748 00:39:38,870 --> 00:39:41,680 to a sugar if it's lactose. 749 00:39:41,680 --> 00:39:46,690 And they show that for the evolved strain, 750 00:39:46,690 --> 00:39:48,730 not only does it no longer have that arrow 751 00:39:48,730 --> 00:39:51,280 in the sense of the response, but it also no longer 752 00:39:51,280 --> 00:39:52,100 has the benefit. 753 00:39:52,100 --> 00:39:53,850 It's good to do both of those measurements 754 00:39:53,850 --> 00:39:56,583 because it's possible you might have missed something. 755 00:39:56,583 --> 00:39:57,083 Yeah. 756 00:39:57,083 --> 00:40:00,265 AUDIENCE: So but like, it's for one protocol going from lactose 757 00:40:00,265 --> 00:40:01,090 to maltose. 758 00:40:01,090 --> 00:40:02,090 PROFESSOR: That's right. 759 00:40:04,800 --> 00:40:05,940 Yep. 760 00:40:05,940 --> 00:40:09,532 And this is the nature of science, 761 00:40:09,532 --> 00:40:11,490 or in particular this kind of science, I think, 762 00:40:11,490 --> 00:40:17,150 that you can collect data that is 763 00:40:17,150 --> 00:40:19,380 consistent with this interpretation, 764 00:40:19,380 --> 00:40:24,020 but it does not at all prove that this is not gut, either. 765 00:40:27,550 --> 00:40:30,260 And so you could have imagined experiments 766 00:40:30,260 --> 00:40:33,230 where you maybe mixed two strains that 767 00:40:33,230 --> 00:40:36,270 follow this direct versus the AAR, 768 00:40:36,270 --> 00:40:39,230 and you mix them together 50-50 in the gut. 769 00:40:39,230 --> 00:40:41,310 And then you go and you do sequencing later. 770 00:40:41,310 --> 00:40:43,690 And actually, people do such kinds of experiments, 771 00:40:43,690 --> 00:40:48,820 and in some cases, amazingly, sometimes something 772 00:40:48,820 --> 00:40:52,562 comes out the other end that is interesting. 773 00:40:52,562 --> 00:40:53,520 I couldn't help myself. 774 00:40:53,520 --> 00:40:54,228 Sorry about that. 775 00:40:57,210 --> 00:40:59,040 So yes. 776 00:40:59,040 --> 00:41:03,270 One other point on this paper, or this figure, 777 00:41:03,270 --> 00:41:06,390 that they even color code the data points. 778 00:41:06,390 --> 00:41:08,880 So they have their main thing in this black, 779 00:41:08,880 --> 00:41:13,190 but then the other ones are red, yellow, blue. 780 00:41:13,190 --> 00:41:18,480 it's telling us that this thing is directional. 781 00:41:18,480 --> 00:41:23,160 It's S1, S2, S2, S1, specificity, extinction. 782 00:41:23,160 --> 00:41:25,790 So they are even kind of telling us 783 00:41:25,790 --> 00:41:28,190 what each of those data points mean. 784 00:41:28,190 --> 00:41:30,570 And you might say, oh, that's kind of overkill. 785 00:41:30,570 --> 00:41:33,660 But given that it's just so important that you're 786 00:41:33,660 --> 00:41:37,051 keeping track of why we're doing each of these measurements 787 00:41:37,051 --> 00:41:39,300 and what it's supposed to tell us about the big story, 788 00:41:39,300 --> 00:41:43,810 I think that it's really just wonderful to be reminded, 789 00:41:43,810 --> 00:41:51,410 OK, what is this controlling for, or what not. 790 00:41:51,410 --> 00:41:52,648 Yes. 791 00:41:52,648 --> 00:41:53,680 AUDIENCE: I like the colors too, but they don't stand out 792 00:41:53,680 --> 00:41:54,888 very well in black and white. 793 00:41:57,240 --> 00:41:59,040 PROFESSOR: Yes, well, OK. 794 00:41:59,040 --> 00:42:02,930 So we can lead entire discussions about both 795 00:42:02,930 --> 00:42:05,050 how to write a paper and how to read a paper. 796 00:42:05,050 --> 00:42:07,410 How to read a paper is that you should just 797 00:42:07,410 --> 00:42:08,960 get a color printer, OK? 798 00:42:08,960 --> 00:42:10,460 So that's the advice. 799 00:42:10,460 --> 00:42:11,890 In terms of how to write a paper, 800 00:42:11,890 --> 00:42:13,790 I would say, though, that it's always 801 00:42:13,790 --> 00:42:17,420 good-- you should take advantage of colors, 802 00:42:17,420 --> 00:42:20,740 but you should choose the color scheme/scale such 803 00:42:20,740 --> 00:42:24,557 that it is legible or compelling in both color 804 00:42:24,557 --> 00:42:25,390 and black and white. 805 00:42:25,390 --> 00:42:27,440 And actually there's a very nice set 806 00:42:27,440 --> 00:42:34,339 of articles written by Wong, who's a graphic artist that 807 00:42:34,339 --> 00:42:35,130 works at the Broad. 808 00:42:35,130 --> 00:42:39,174 He wrote maybe 15 different kind of one- or two-page articles 809 00:42:39,174 --> 00:42:40,590 for Nature Methods, basically just 810 00:42:40,590 --> 00:42:41,714 about how to write figures. 811 00:42:41,714 --> 00:42:43,800 And he talks about these different color scales. 812 00:42:43,800 --> 00:42:47,990 You know, there's-- what do the words mean when 813 00:42:47,990 --> 00:42:50,670 there's brightness? 814 00:42:50,670 --> 00:42:51,844 Color, hue, saturation. 815 00:42:51,844 --> 00:42:54,010 I can't remember what all these things mean anymore. 816 00:42:54,010 --> 00:42:56,700 But that's why you should read the articles. 817 00:42:56,700 --> 00:42:59,456 And it's fascinating stuff, actually. 818 00:43:06,820 --> 00:43:10,120 So this is the basic argument they made in this paper, 819 00:43:10,120 --> 00:43:14,830 that because E. coli are exposed to a typical order of carbon 820 00:43:14,830 --> 00:43:19,510 sources in the mammalian gut, that that's 821 00:43:19,510 --> 00:43:22,375 the possible explanation for why this thing evolved that they 822 00:43:22,375 --> 00:43:23,750 can measure experimentally, which 823 00:43:23,750 --> 00:43:26,890 is that when, these E. coli, they see lactose, 824 00:43:26,890 --> 00:43:30,000 they start activating the mall genes but not vice versa. 825 00:43:30,000 --> 00:43:33,720 Now, like always, you can't prove 826 00:43:33,720 --> 00:43:35,280 that this is what's going on. 827 00:43:35,280 --> 00:43:37,740 I think very interesting set of ideas. 828 00:43:37,740 --> 00:43:41,640 Some of the people that study gut microbes just 829 00:43:41,640 --> 00:43:45,390 think that it's implausible that this would actually work out 830 00:43:45,390 --> 00:43:47,920 in the sense that the fitness cost benefits, et cetera, 831 00:43:47,920 --> 00:43:49,704 would be relevant. 832 00:43:49,704 --> 00:43:51,620 Do we know that the time scales even work out? 833 00:43:51,620 --> 00:43:54,200 I don't know how-- so there are a bunch of things 834 00:43:54,200 --> 00:43:56,350 that you could worry about. 835 00:43:56,350 --> 00:43:58,320 And I would say I don't know enough about what 836 00:43:58,320 --> 00:44:01,580 goes on in the gut to actually have an educated 837 00:44:01,580 --> 00:44:02,680 opinion on that point. 838 00:44:02,680 --> 00:44:08,200 But I think it's a very kind of clear exposition of the ideas. 839 00:44:08,200 --> 00:44:09,676 Yeah. 840 00:44:09,676 --> 00:44:11,644 AUDIENCE: Do you have a sense for what 841 00:44:11,644 --> 00:44:19,024 fraction of E. coli bacteria are in the gut of animals? 842 00:44:19,024 --> 00:44:20,008 Do you see what I mean? 843 00:44:20,008 --> 00:44:22,000 Because it's just a very small fraction. 844 00:44:22,000 --> 00:44:23,000 PROFESSOR: That's right. 845 00:44:23,000 --> 00:44:26,930 Yeah, and I guess-- so I myself definitely 846 00:44:26,930 --> 00:44:29,470 do not know what fraction of E.coli are in the gut as 847 00:44:29,470 --> 00:44:31,712 compared to in the soil or in the-- 848 00:44:31,712 --> 00:44:34,170 it's not that E. coli are the dominant bacteria in our gut, 849 00:44:34,170 --> 00:44:34,730 either. 850 00:44:34,730 --> 00:44:39,250 It's just that they do enter the gut at some rate. 851 00:44:39,250 --> 00:44:42,670 And then the question is, if we knew the answer, 852 00:44:42,670 --> 00:44:45,852 I don't even know if it would tell us-- I mean, 853 00:44:45,852 --> 00:44:46,810 it gives some guidance. 854 00:44:46,810 --> 00:44:48,470 But there's also a question of where 855 00:44:48,470 --> 00:44:50,220 is it that cells are actually dividing, 856 00:44:50,220 --> 00:44:52,610 because it could be the case that the life cycle is 857 00:44:52,610 --> 00:44:54,370 that they go in the gut, and that's where 858 00:44:54,370 --> 00:44:55,540 they're dividing and happy. 859 00:44:55,540 --> 00:44:58,560 And then they end up in some soil somewhere, 860 00:44:58,560 --> 00:45:02,000 and they're in some stationary phase, and maybe most of them 861 00:45:02,000 --> 00:45:03,870 are dying at some small rate. 862 00:45:03,870 --> 00:45:05,690 Eventually some of them get uptaken again 863 00:45:05,690 --> 00:45:07,231 and then they divide, divide, divide. 864 00:45:07,231 --> 00:45:10,730 So it could be the cases that only 0.1% of all E. coli, 865 00:45:10,730 --> 00:45:11,980 or 01% of the time. 866 00:45:11,980 --> 00:45:13,520 It could be a very small amount of time or something, 867 00:45:13,520 --> 00:45:15,269 and they're in the gut, but it could still 868 00:45:15,269 --> 00:45:20,610 be the dominant event in their lives. 869 00:45:20,610 --> 00:45:21,667 So I just don't know. 870 00:45:25,010 --> 00:45:29,190 I think I'm maybe going to skip the second half of the talk, 871 00:45:29,190 --> 00:45:31,480 or second half of the paper, where they discuss 872 00:45:31,480 --> 00:45:37,930 this idea of the typical order of events in which yeast 873 00:45:37,930 --> 00:45:41,380 in the context of fermentation and making wine 874 00:45:41,380 --> 00:45:44,231 might encounter or experience different environmental 875 00:45:44,231 --> 00:45:44,730 assaults. 876 00:45:44,730 --> 00:45:46,854 And it's a similar kind of idea of cross-protection 877 00:45:46,854 --> 00:45:49,405 against different assaults. 878 00:45:49,405 --> 00:45:52,070 But in some ways, this is really sort 879 00:45:52,070 --> 00:45:56,210 of two mini-papers that tie together 880 00:45:56,210 --> 00:46:00,480 in this general thing of anticipatory regulation. 881 00:46:00,480 --> 00:46:04,540 Are there any questions about this before we move on? 882 00:46:08,977 --> 00:46:09,963 Yes. 883 00:46:09,963 --> 00:46:13,003 AUDIENCE: I have a question about when 884 00:46:13,003 --> 00:46:15,206 the baby males grow up. 885 00:46:15,206 --> 00:46:16,372 Do they need this mechanism? 886 00:46:19,330 --> 00:46:22,207 Do E. coli? 887 00:46:22,207 --> 00:46:22,790 PROFESSOR: OK. 888 00:46:22,790 --> 00:46:27,310 So your question is whether E. coli in the gut of adults 889 00:46:27,310 --> 00:46:30,280 have this. 890 00:46:30,280 --> 00:46:37,720 I don't know, but my feeling is that these things are probably 891 00:46:37,720 --> 00:46:42,220 fairly transient in the sense that you probably 892 00:46:42,220 --> 00:46:47,180 don't have-- well, certainly, from the standpoint of entering 893 00:46:47,180 --> 00:46:49,460 the gut, the bacteria don't have any choice 894 00:46:49,460 --> 00:46:50,620 about whose gut they enter. 895 00:46:50,620 --> 00:46:52,870 So it probably is going to be the case that they 896 00:46:52,870 --> 00:46:54,560 have to have a basic strategy that 897 00:46:54,560 --> 00:46:58,850 is going to work across both adults and the kids, 898 00:46:58,850 --> 00:47:01,890 although it's true that the gut communities in kids 899 00:47:01,890 --> 00:47:03,550 are different from adults. 900 00:47:03,550 --> 00:47:06,037 But it's hard to know what to make of that. 901 00:47:06,037 --> 00:47:06,975 Yeah. 902 00:47:06,975 --> 00:47:12,300 AUDIENCE: So here they decouple the two responses. 903 00:47:12,300 --> 00:47:16,644 Are there any examples of researchers coupling them, 904 00:47:16,644 --> 00:47:17,550 like Pavlov's? 905 00:47:17,550 --> 00:47:18,670 PROFESSOR: Yeah. 906 00:47:18,670 --> 00:47:20,258 That's a good question. 907 00:47:20,258 --> 00:47:22,216 I know that I've heard people talking about it, 908 00:47:22,216 --> 00:47:24,300 but I can't point to a specific example 909 00:47:24,300 --> 00:47:27,883 of where it's been convincing. 910 00:47:27,883 --> 00:47:29,758 That doesn't mean it's not out there, though. 911 00:47:33,620 --> 00:47:35,686 It's a very important, interesting question. 912 00:47:43,284 --> 00:47:45,200 So what we want to now is kind of switch gears 913 00:47:45,200 --> 00:47:48,984 a little bit and switch to this topic of-- it's 914 00:47:48,984 --> 00:47:51,400 going to be kind of partly about phenotypic heterogeneity, 915 00:47:51,400 --> 00:47:54,660 but partly again about fluctuating environments. 916 00:47:54,660 --> 00:47:57,010 And the way that those two topics are related 917 00:47:57,010 --> 00:47:58,910 is via bet hedging, but I just want 918 00:47:58,910 --> 00:48:04,470 to highlight that bet hedging is a way of using 919 00:48:04,470 --> 00:48:08,220 phenotypic heterogeneity in order to survive fluctuating 920 00:48:08,220 --> 00:48:12,024 environments, whereas it's not the only explanation 921 00:48:12,024 --> 00:48:13,440 for this phenotypic heterogeneity. 922 00:48:24,040 --> 00:48:34,940 So yeah, I think maybe we'll-- so first of all, 923 00:48:34,940 --> 00:48:39,750 on the topic of phenotypic heterogeneity, 924 00:48:39,750 --> 00:48:44,260 when we say that we are often referring to the fact that even 925 00:48:44,260 --> 00:48:47,430 clonal populations can be remarkably heterogeneous 926 00:48:47,430 --> 00:48:48,710 in a given environment. 927 00:48:48,710 --> 00:48:50,480 So we'll say phenotypic heterogeneity 928 00:48:50,480 --> 00:48:55,397 in clonal populations, because of course, 929 00:48:55,397 --> 00:48:57,730 if there's a lot of genetic diversity in the population, 930 00:48:57,730 --> 00:48:59,934 then it's not, perhaps, a surprise there would 931 00:48:59,934 --> 00:49:01,100 be phenotypic heterogeneity. 932 00:49:05,050 --> 00:49:07,740 What we're going to do is go through these three 933 00:49:07,740 --> 00:49:11,840 different possible explanations and discuss, maybe, 934 00:49:11,840 --> 00:49:13,990 canonical biological examples of each. 935 00:49:21,920 --> 00:49:25,700 And the examples that we're going to give-- well, 936 00:49:25,700 --> 00:49:28,760 the classic example is in the context of seed germination. 937 00:49:37,870 --> 00:49:40,520 But more recently there's this question 938 00:49:40,520 --> 00:49:45,470 of the persister cells, or the persistence phenotype. 939 00:49:52,760 --> 00:49:56,280 Now in both of these cases, it's really 940 00:49:56,280 --> 00:49:59,810 a way in which a clonal population can 941 00:49:59,810 --> 00:50:03,700 send a small population of either seeds or cells 942 00:50:03,700 --> 00:50:08,780 into a protected state that's protected against some assault. 943 00:50:08,780 --> 00:50:11,640 So seed germination, this is a case 944 00:50:11,640 --> 00:50:14,250 where the seed, if it doesn't germinate, 945 00:50:14,250 --> 00:50:17,646 it's protected against draughts. 946 00:50:17,646 --> 00:50:18,146 Draught. 947 00:50:23,366 --> 00:50:24,865 And this one, maybe, is antibiotics. 948 00:50:30,182 --> 00:50:31,640 What I want to do is just make sure 949 00:50:31,640 --> 00:50:36,140 that we understand the seed germination idea. 950 00:50:36,140 --> 00:50:39,010 And so this is, from the standpoint of ecology, 951 00:50:39,010 --> 00:50:43,580 these ideas came out in the '70s, at least. 952 00:50:43,580 --> 00:50:48,520 So what we can imagine is a situation where you are some, 953 00:50:48,520 --> 00:50:49,550 maybe, desert annual. 954 00:50:53,120 --> 00:50:57,310 And so you send out these seeds. 955 00:50:57,310 --> 00:51:01,030 Now most of the time, you get enough rain, 956 00:51:01,030 --> 00:51:02,240 the seeds are fine. 957 00:51:02,240 --> 00:51:05,430 They can germinate, sprout the little seedlings, 958 00:51:05,430 --> 00:51:08,370 and they'll be able to grow up just fine. 959 00:51:08,370 --> 00:51:10,830 But at some rate there might be a severe drought 960 00:51:10,830 --> 00:51:14,600 that would kill all the cells that actually germinated. 961 00:51:14,600 --> 00:51:19,130 So the question is, what might be the optimal strategy? 962 00:51:19,130 --> 00:51:20,730 So for example, let's just imagine 963 00:51:20,730 --> 00:51:27,840 that we live in a world in which 75% of the time 964 00:51:27,840 --> 00:51:34,450 there's rain, 25% of the time there's no rain. 965 00:51:37,230 --> 00:51:44,233 Now if you germinate-- so a particular seed, 966 00:51:44,233 --> 00:51:49,070 if it germinates, it pops out-- then 967 00:51:49,070 --> 00:51:55,810 it will expect to send two seeds to the following year, 968 00:51:55,810 --> 00:51:59,380 whereas as it germinates and there's no rain, 969 00:51:59,380 --> 00:52:05,380 then it only has a 10% probability of surviving, 970 00:52:05,380 --> 00:52:10,560 or that it could on average send 1/10 of a seed 971 00:52:10,560 --> 00:52:13,910 to the next generation. 972 00:52:13,910 --> 00:52:15,760 Now of course, you don't have to germinate. 973 00:52:15,760 --> 00:52:22,230 If you don't germinate, then what happens 974 00:52:22,230 --> 00:52:28,770 is that you just send one seed to the next year. 975 00:52:28,770 --> 00:52:31,190 If you'd like, you can include some probability 976 00:52:31,190 --> 00:52:34,411 that the seed doesn't survive, but for simplicity, we 977 00:52:34,411 --> 00:52:34,910 can do that. 978 00:52:38,090 --> 00:52:40,840 So the idea is that you don't know whether it's going to rain 979 00:52:40,840 --> 00:52:41,890 or not. 980 00:52:41,890 --> 00:52:44,540 But what you can do is you can either germinate or not 981 00:52:44,540 --> 00:52:46,760 germinate. 982 00:52:46,760 --> 00:52:49,206 And then this tells us how many seeds make it 983 00:52:49,206 --> 00:52:50,510 to the following year. 984 00:52:50,510 --> 00:52:54,530 Do you understand the framework? 985 00:52:54,530 --> 00:53:00,580 So the question is, which one of these strategies 986 00:53:00,580 --> 00:53:03,970 has a higher long term growth rate 987 00:53:03,970 --> 00:53:07,257 in terms of the number of seeds that you 988 00:53:07,257 --> 00:53:08,215 have in the population? 989 00:53:21,220 --> 00:53:23,950 I'm going to give you 30 seconds just 990 00:53:23,950 --> 00:53:25,450 to kind of think about this, play 991 00:53:25,450 --> 00:53:27,130 with the numbers a little bit. 992 00:53:27,130 --> 00:53:29,570 And the question is, should you just 993 00:53:29,570 --> 00:53:32,054 have a strategy where you germinate or don't germinate? 994 00:53:32,054 --> 00:53:33,470 And then what we're going to find, 995 00:53:33,470 --> 00:53:35,950 maybe, is that the optimal strategy is 996 00:53:35,950 --> 00:53:38,464 to do this probabilistically. 997 00:53:38,464 --> 00:53:40,380 But it's useful to just play with such numbers 998 00:53:40,380 --> 00:53:41,088 for a little bit. 999 00:54:11,840 --> 00:54:13,590 The physics office, unfortunately, 1000 00:54:13,590 --> 00:54:16,310 is trying to steal our cards from us, so we don't have them. 1001 00:54:16,310 --> 00:54:18,010 But what we're going to instead do 1002 00:54:18,010 --> 00:54:20,939 is, if you think you want to germinate 1003 00:54:20,939 --> 00:54:21,980 you raise your left hand. 1004 00:54:21,980 --> 00:54:24,229 If you don't want to germinate, raise your right hand. 1005 00:54:24,229 --> 00:54:24,770 OK. 1006 00:54:24,770 --> 00:54:29,460 Ready, 3, 2, 1. 1007 00:54:29,460 --> 00:54:33,710 So we have a fair distribution of-- it really 1008 00:54:33,710 --> 00:54:35,049 is kind of 50-50. 1009 00:54:35,049 --> 00:54:35,840 That's interesting. 1010 00:54:35,840 --> 00:54:41,106 All right, let's go ahead and-- yeah, let's spend a minute. 1011 00:54:41,106 --> 00:54:43,730 Turn to your neighbor and try to convince them that it's either 1012 00:54:43,730 --> 00:54:44,854 better to germinate or not. 1013 00:54:44,854 --> 00:55:23,316 [SIDE CONVERSATIONS] 1014 00:55:23,316 --> 00:55:28,601 You can start with 100 of one, 100 of another, and then just-- 1015 00:55:28,601 --> 00:55:30,559 AUDIENCE: Start with 100 and then you just go-- 1016 00:55:30,559 --> 00:55:40,780 [SIDE CONVERSATIONS] 1017 00:55:40,780 --> 00:55:43,030 PROFESSOR: Let's say 10 to the 4. 1018 00:55:43,030 --> 00:55:44,885 So we're trying to not think about this-- 1019 00:55:44,885 --> 00:56:06,067 [SIDE CONVERSATIONS] 1020 00:56:06,067 --> 00:56:07,900 So there's another question about short term 1021 00:56:07,900 --> 00:56:08,810 versus long term. 1022 00:56:08,810 --> 00:56:12,310 And we're actually trying to not worry about that. 1023 00:56:12,310 --> 00:56:14,917 So you can think about large populations, so it's really-- 1024 00:56:14,917 --> 00:56:16,292 AUDIENCE: This is after one year. 1025 00:56:16,292 --> 00:56:18,220 You will maximize your expected value. 1026 00:56:18,220 --> 00:56:19,340 PROFESSOR: That's why I said the long term 1027 00:56:19,340 --> 00:56:20,381 growth of the population. 1028 00:56:20,381 --> 00:56:25,795 So if we just compare-- so why don't we reconvene. 1029 00:56:25,795 --> 00:56:27,920 Various people have been talking about the question 1030 00:56:27,920 --> 00:56:30,564 of how many seeds you're starting with, 1031 00:56:30,564 --> 00:56:32,230 what the population size is, because you 1032 00:56:32,230 --> 00:56:33,560 have to worry about extinction. 1033 00:56:33,560 --> 00:56:35,518 And that's another kind of interesting question 1034 00:56:35,518 --> 00:56:37,752 of short term versus long term payouts of fitness. 1035 00:56:37,752 --> 00:56:39,460 That's actually not the effect that we're 1036 00:56:39,460 --> 00:56:40,930 trying to get at right now. 1037 00:56:40,930 --> 00:56:43,190 So we can even think about this in the limit of just 1038 00:56:43,190 --> 00:56:44,420 large population sizes. 1039 00:56:44,420 --> 00:56:47,650 And we want to compare the two strategies. 1040 00:56:47,650 --> 00:56:49,599 And for any strategy you choose, including 1041 00:56:49,599 --> 00:56:51,140 the probabilistic strategies, there's 1042 00:56:51,140 --> 00:56:54,120 going to be a long term growth rate of the population. 1043 00:56:54,120 --> 00:56:56,882 So it will grow or shrink exponentially over time. 1044 00:56:56,882 --> 00:56:58,340 And what you want to do is you want 1045 00:56:58,340 --> 00:57:00,506 to find the strategy, in the context of bet hedging, 1046 00:57:00,506 --> 00:57:03,831 you want to find the strategy that maximizes this long term 1047 00:57:03,831 --> 00:57:05,080 growth rate of the population. 1048 00:57:05,080 --> 00:57:09,020 And this is essentially what Edo talks about in his paper, 1049 00:57:09,020 --> 00:57:12,190 but in a more mathematical, general-- because he's talking 1050 00:57:12,190 --> 00:57:16,510 about n different phenotypes and environments 1051 00:57:16,510 --> 00:57:17,790 and arbitrary switching rates. 1052 00:57:17,790 --> 00:57:21,090 And it's beautiful and general and he makes nice connections 1053 00:57:21,090 --> 00:57:22,650 to information theory. 1054 00:57:22,650 --> 00:57:25,330 But it's possible to read that paper and then not 1055 00:57:25,330 --> 00:57:28,090 actually understand what bet hedging, 1056 00:57:28,090 --> 00:57:29,840 or what any of these things are, quite. 1057 00:57:29,840 --> 00:57:32,050 So it's good to take the simple thing, 1058 00:57:32,050 --> 00:57:34,780 and then it's also good to know that somebody like Edo 1059 00:57:34,780 --> 00:57:36,196 has thought about this very deeply 1060 00:57:36,196 --> 00:57:38,280 and figured out all these beautiful things. 1061 00:57:38,280 --> 00:57:42,280 And I know that paper was hard, so don't-- but it's really like 1062 00:57:42,280 --> 00:57:45,580 a two-page paper if you look at the text. 1063 00:57:45,580 --> 00:57:47,410 So it shouldn't take too much time just 1064 00:57:47,410 --> 00:57:49,201 to get a sense of what he's thinking about. 1065 00:57:49,201 --> 00:57:52,570 But then it's good, also, to be concrete in a simple situation. 1066 00:57:52,570 --> 00:57:55,052 So what we want to know is that, over the long term-- 1067 00:57:55,052 --> 00:57:56,760 then what you can do is you can just say, 1068 00:57:56,760 --> 00:58:02,290 well, the ratios of rain, no rain over the long term 1069 00:58:02,290 --> 00:58:06,100 will be 3 to 1, because it's over a long term. 1070 00:58:06,100 --> 00:58:07,960 It doesn't matter what order they come in 1071 00:58:07,960 --> 00:58:10,316 because over the long run it's all going to average out. 1072 00:58:10,316 --> 00:58:12,190 We're assuming we're a large population size, 1073 00:58:12,190 --> 00:58:14,480 so we don't have to worry about stochastic extinction. 1074 00:58:14,480 --> 00:58:15,980 The question is, since we don't have 1075 00:58:15,980 --> 00:58:18,040 to worry about stochastic extinction, 1076 00:58:18,040 --> 00:58:22,450 then can we just say that we just 1077 00:58:22,450 --> 00:58:25,660 calculate, for example, the mean weighted 1078 00:58:25,660 --> 00:58:28,120 by these fractions of these numbers? 1079 00:58:28,120 --> 00:58:30,710 And actually, the subtle possible thing 1080 00:58:30,710 --> 00:58:34,350 is that it's the geometric mean that is relevant, 1081 00:58:34,350 --> 00:58:35,480 not the normal mean. 1082 00:58:38,070 --> 00:58:41,340 And that's somehow tricky when given these numbers. 1083 00:58:41,340 --> 00:58:46,760 But it's sort of obvious if you just switch this to a 0. 1084 00:58:46,760 --> 00:58:49,364 Because if this is a 0, then is germinate 1085 00:58:49,364 --> 00:58:50,530 going to be a good strategy? 1086 00:58:53,050 --> 00:58:54,100 No, right? 1087 00:58:54,100 --> 00:58:58,730 Because what that would mean is that if everybody always 1088 00:58:58,730 --> 00:59:01,770 germinates, it may be that this number could be 1,000, 1089 00:59:01,770 --> 00:59:05,560 and it's wonderful except that if, at some low rate, 1090 00:59:05,560 --> 00:59:09,260 there's no rain and this is a 0, then the population's dead. 1091 00:59:09,260 --> 00:59:11,353 And you'll always encounter that eventually. 1092 00:59:11,353 --> 00:59:12,727 So that just highlights that it's 1093 00:59:12,727 --> 00:59:15,820 the geometric mean you want to be focusing on. 1094 00:59:15,820 --> 00:59:18,910 And in particular, we know that on average, over the long term, 1095 00:59:18,910 --> 00:59:20,160 there are going to be three of these years for each 1096 00:59:20,160 --> 00:59:20,820 of these years. 1097 00:59:20,820 --> 00:59:22,710 So we can just multiply the numbers. 1098 00:59:22,710 --> 00:59:26,890 So it's really going to be there's going to be three times 1099 00:59:26,890 --> 00:59:33,770 where you double for each one time that you divide by 10. 1100 00:59:33,770 --> 00:59:36,480 What you see is that over the long run, 1101 00:59:36,480 --> 00:59:44,960 this population shrinks, whereas in this case it's 1 times 1102 00:59:44,960 --> 00:59:47,030 1 times 1 times 1. 1103 00:59:49,800 --> 00:59:52,350 So the long term growth rate/shrinking 1104 00:59:52,350 --> 00:59:54,880 rate of these populations is that in the long run 1105 00:59:54,880 --> 00:59:56,675 this population does not change in size, 1106 00:59:56,675 --> 00:59:59,540 whereas over the long run, this population shrinks. 1107 00:59:59,540 --> 01:00:01,790 AUDIENCE: So in the geometric mean the expected change 1108 01:00:01,790 --> 01:00:04,922 in the log population, which is also the expected percentage 1109 01:00:04,922 --> 01:00:05,963 change in the population. 1110 01:00:05,963 --> 01:00:06,546 Is that right? 1111 01:00:09,755 --> 01:00:10,380 PROFESSOR: Yes. 1112 01:00:10,380 --> 01:00:11,730 So I think that if you take the logs, 1113 01:00:11,730 --> 01:00:12,730 it ends up being equivalent. 1114 01:00:12,730 --> 01:00:14,605 But you have to make sure you just do it all. 1115 01:00:18,850 --> 01:00:21,650 And I guess what I want to highlight in this case is that, 1116 01:00:21,650 --> 01:00:25,310 A, you have to be careful of taking averages, but B, 1117 01:00:25,310 --> 01:00:27,090 to get the intuition behind the situations 1118 01:00:27,090 --> 01:00:29,480 where bet hedging, I think, is really valuable, 1119 01:00:29,480 --> 01:00:34,550 it's situations where this thing gets very small 1120 01:00:34,550 --> 01:00:36,610 and this might become rather large. 1121 01:00:36,610 --> 01:00:39,810 So you can imagine that this could be 200 1122 01:00:39,810 --> 01:00:42,310 and this could be 10 to the minus 9. 1123 01:00:46,430 --> 01:00:48,790 Or if you'd like, you could say that's 0. 1124 01:00:48,790 --> 01:00:52,160 And it's clear that if this is 0, then 1125 01:00:52,160 --> 01:00:53,590 you can't germinate all the time, 1126 01:00:53,590 --> 01:00:55,256 because then the population's definitely 1127 01:00:55,256 --> 01:00:56,870 going to go extinct over the long run. 1128 01:00:56,870 --> 01:00:58,290 But this is also a situation where 1129 01:00:58,290 --> 01:01:01,290 there's real tension between long term and short term 1130 01:01:01,290 --> 01:01:05,840 situations if this probability is very small. 1131 01:01:05,840 --> 01:01:08,470 But it's in these situations that it 1132 01:01:08,470 --> 01:01:12,280 pays, in terms of the long run growth rate of the population, 1133 01:01:12,280 --> 01:01:16,366 to do this some fraction of the time. 1134 01:01:16,366 --> 01:01:17,490 And it could be very small. 1135 01:01:17,490 --> 01:01:21,960 It could be that only one in 10 to the 4 of the cells 1136 01:01:21,960 --> 01:01:25,920 enter this persister state that is resistant to antibiotics. 1137 01:01:25,920 --> 01:01:27,991 But it can maximize the long term growth 1138 01:01:27,991 --> 01:01:28,949 rate of the population. 1139 01:01:33,530 --> 01:01:35,952 There are various models that you 1140 01:01:35,952 --> 01:01:38,410 can write down for either seed germination or this or that. 1141 01:01:38,410 --> 01:01:43,110 But the relevant thing is, for example, in the persister case, 1142 01:01:43,110 --> 01:01:46,760 it might be that the rate of persister formation 1143 01:01:46,760 --> 01:01:49,070 might only be one cell in 10 to the 5 or 10 to the 6. 1144 01:01:49,070 --> 01:01:51,430 So a very small sub-population of the cells 1145 01:01:51,430 --> 01:01:53,760 are in this so-called persister state. 1146 01:01:53,760 --> 01:01:56,510 But those cells, they have a couple of properties. 1147 01:01:56,510 --> 01:01:59,280 One is that they are dividing slowly. 1148 01:01:59,280 --> 01:02:02,379 That means there really is a cost to entering that persister 1149 01:02:02,379 --> 01:02:04,170 state, but if it's only one of 10 to the 5, 1150 01:02:04,170 --> 01:02:07,370 that means that it's really a rather small cost in terms 1151 01:02:07,370 --> 01:02:10,490 of decreasing the growth of the population. 1152 01:02:10,490 --> 01:02:13,195 And those cells also could be resistant to antibiotics. 1153 01:02:13,195 --> 01:02:14,820 So you can select for them in the sense 1154 01:02:14,820 --> 01:02:16,278 that you add a bunch of antibiotics 1155 01:02:16,278 --> 01:02:21,360 and maybe only the persister cells are left after 12 hours. 1156 01:02:21,360 --> 01:02:25,580 But even though I say you selected for them, 1157 01:02:25,580 --> 01:02:29,350 it's not a true antibiotic resistance kind of mutation, 1158 01:02:29,350 --> 01:02:31,320 because if you take that persister population, 1159 01:02:31,320 --> 01:02:32,890 you let it grow back up, what you 1160 01:02:32,890 --> 01:02:35,949 find is that it is still sensitive to the antibiotic. 1161 01:02:35,949 --> 01:02:38,240 Now it's also possible that at some low rate, maybe one 1162 01:02:38,240 --> 01:02:41,150 in 10 to the 8, you evolve resistance to an antibiotic. 1163 01:02:41,150 --> 01:02:44,010 That's genetic resistance that stays with you, 1164 01:02:44,010 --> 01:02:46,950 whereas the persister kind of resistance, this 1165 01:02:46,950 --> 01:02:50,750 is something that is a phenotypic switch. 1166 01:02:50,750 --> 01:02:52,970 And when it grows back up, it kind of 1167 01:02:52,970 --> 01:02:55,490 reverts back to the original phenotype, 1168 01:02:55,490 --> 01:02:58,170 which is being sensitive to the antibiotic. 1169 01:02:58,170 --> 01:03:00,380 So that's the classic way that you distinguish 1170 01:03:00,380 --> 01:03:04,114 between genetic resistance and this persister-based or 1171 01:03:04,114 --> 01:03:05,030 phenotypic resistance. 1172 01:03:05,030 --> 01:03:05,380 Yes. 1173 01:03:05,380 --> 01:03:06,921 AUDIENCE: What would be the mechanism 1174 01:03:06,921 --> 01:03:08,140 of something like that? 1175 01:03:08,140 --> 01:03:09,139 PROFESSOR: That's right. 1176 01:03:09,139 --> 01:03:11,141 So people argue a fair amount about this. 1177 01:03:11,141 --> 01:03:12,640 And there's a thought that there are 1178 01:03:12,640 --> 01:03:19,330 these toxin, anti-toxin modules in bacteria that are normally 1179 01:03:19,330 --> 01:03:21,070 suppressed but can be triggered, and then 1180 01:03:21,070 --> 01:03:22,880 can lead to these states entering, 1181 01:03:22,880 --> 01:03:28,900 although this ends up kind of a muddy debate/subject somehow. 1182 01:03:28,900 --> 01:03:32,480 But the idea is that there's something in the cell that 1183 01:03:32,480 --> 01:03:36,290 is normally repressed, but then at some rate 1184 01:03:36,290 --> 01:03:39,304 activates and causes the cell to-- and of course, 1185 01:03:39,304 --> 01:03:41,970 you can always argue, OK, did it evolve for this purpose or not? 1186 01:03:41,970 --> 01:03:44,320 Because also, if you're dividing slowly, 1187 01:03:44,320 --> 01:03:47,170 then you're going to be resistant to many things. 1188 01:03:47,170 --> 01:03:49,580 But then there's also arguments that there are somehow 1189 01:03:49,580 --> 01:03:51,520 more specific persister-type states 1190 01:03:51,520 --> 01:03:53,810 against particular assaults or particular antibiotics. 1191 01:03:53,810 --> 01:03:56,485 So I think that there are many subtleties 1192 01:03:56,485 --> 01:03:57,360 to this whole debate. 1193 01:04:07,850 --> 01:04:11,010 Any other questions about this idea of bet hedging? 1194 01:04:15,150 --> 01:04:18,290 So I think that bet hedging is indeed 1195 01:04:18,290 --> 01:04:23,290 one possible way in which evolution 1196 01:04:23,290 --> 01:04:26,520 could select for phenotypic heterogeneity in a population. 1197 01:04:26,520 --> 01:04:29,620 So this is a population of seeds that are all, in principle, 1198 01:04:29,620 --> 01:04:30,480 identical. 1199 01:04:30,480 --> 01:04:32,500 You put them out and you plant them all 1200 01:04:32,500 --> 01:04:34,320 and they see the same environment. 1201 01:04:34,320 --> 01:04:36,230 They don't all germinate. 1202 01:04:36,230 --> 01:04:38,560 But it's not that all those seeds that didn't germinate 1203 01:04:38,560 --> 01:04:41,760 are dead, because actually, the following year, they still 1204 01:04:41,760 --> 01:04:44,206 go back and they can follow some stochastic strategy. 1205 01:04:44,206 --> 01:04:46,080 So I just want to stress that right now we're 1206 01:04:46,080 --> 01:04:49,260 considering the two extremes, but for bet hedging, 1207 01:04:49,260 --> 01:04:54,117 you say probability germinate, 1 minus p, don't germinate. 1208 01:04:54,117 --> 01:04:56,700 And those are, in general, going to be some probability p that 1209 01:04:56,700 --> 01:04:58,882 can be non-0 that maximizes long term growth 1210 01:04:58,882 --> 01:04:59,840 rate of the population. 1211 01:05:04,100 --> 01:05:08,130 But I do want to talk about two other possible explanations, 1212 01:05:08,130 --> 01:05:12,442 which is this idea of mixed strategies and altruism. 1213 01:05:12,442 --> 01:05:13,900 Now mixed strategies, in some ways, 1214 01:05:13,900 --> 01:05:16,040 is an obvious one, because we just 1215 01:05:16,040 --> 01:05:21,170 got done talking about mixed strategies in game theory, 1216 01:05:21,170 --> 01:05:25,470 where there is this notion that in a given game, 1217 01:05:25,470 --> 01:05:29,990 sometimes this Nash equilibrium was a pure strategy, 1218 01:05:29,990 --> 01:05:31,510 or the evolutionary stable state. 1219 01:05:31,510 --> 01:05:34,880 But sometimes at the equilibrium there 1220 01:05:34,880 --> 01:05:36,671 was coexistence of these two strategies, 1221 01:05:36,671 --> 01:05:38,170 and that could be either implemented 1222 01:05:38,170 --> 01:05:43,369 by two different genotypes, or by one genotype implementing 1223 01:05:43,369 --> 01:05:44,035 both phenotypes. 1224 01:05:47,250 --> 01:05:50,540 So this was the classic game that people talk about 1225 01:05:50,540 --> 01:05:54,580 in this area is this hawk dove game. 1226 01:05:54,580 --> 01:05:58,700 We talked about this just a little bit in class. 1227 01:05:58,700 --> 01:06:00,480 So this is a model of animal conflict. 1228 01:06:11,880 --> 01:06:14,700 And the idea is the animals can either 1229 01:06:14,700 --> 01:06:19,350 be-- so the idea is that there are two animals 1230 01:06:19,350 --> 01:06:21,291 and they come across some resource. 1231 01:06:21,291 --> 01:06:23,040 And the resource could be food or it could 1232 01:06:23,040 --> 01:06:24,640 be a mate or something else. 1233 01:06:24,640 --> 01:06:26,580 And the question is, should those animals 1234 01:06:26,580 --> 01:06:29,000 fight for the resource or should they back down? 1235 01:06:32,090 --> 01:06:33,920 Now the assumption in all this is 1236 01:06:33,920 --> 01:06:36,840 that the hawk is the strategy that fights and the dove 1237 01:06:36,840 --> 01:06:39,690 is the strategy that backs down. 1238 01:06:39,690 --> 01:06:43,050 Now the question is, how do we assign 1239 01:06:43,050 --> 01:06:45,830 some sort of costs and benefits to these various strategies? 1240 01:06:45,830 --> 01:06:47,550 Again, where this is the situation where 1241 01:06:47,550 --> 01:06:49,050 we take the two-player game and then 1242 01:06:49,050 --> 01:06:51,470 we'll kind of generalize it to the population. 1243 01:06:51,470 --> 01:06:54,250 Well, if two doves encounter each other, 1244 01:06:54,250 --> 01:06:57,250 then they don't need to fight at all, 1245 01:06:57,250 --> 01:06:59,340 and they just split the benefit. 1246 01:06:59,340 --> 01:07:01,382 So there's some benefit to the resource, 1247 01:07:01,382 --> 01:07:02,590 and they each get half of it. 1248 01:07:05,600 --> 01:07:08,502 Now of course, if a hawk meets a dove, 1249 01:07:08,502 --> 01:07:13,680 the hawk actually does better, because then the hawk gets 1250 01:07:13,680 --> 01:07:19,220 the entire benefit, doesn't have to pay the cost. 1251 01:07:19,220 --> 01:07:21,670 So already we can see from what I've said so far, 1252 01:07:21,670 --> 01:07:25,150 is dove a Nash equilibrium? 1253 01:07:25,150 --> 01:07:28,590 No, because you can already see that if your opponent is 1254 01:07:28,590 --> 01:07:30,340 a dove, it's better for you. 1255 01:07:30,340 --> 01:07:32,298 If you start in the situation where everybody's 1256 01:07:32,298 --> 01:07:34,380 a dove, so everybody's getting b over 2, 1257 01:07:34,380 --> 01:07:36,480 then one individual has the incentive 1258 01:07:36,480 --> 01:07:40,570 to change strategies and get the full benefit. 1259 01:07:40,570 --> 01:07:44,370 That hawk never has to fight anybody because they will all 1260 01:07:44,370 --> 01:07:46,000 back down. 1261 01:07:46,000 --> 01:07:48,740 Now of course this dove that fought 1262 01:07:48,740 --> 01:07:53,160 that hawk doesn't get any of the resource, 1263 01:07:53,160 --> 01:07:56,190 but also doesn't have to fight. 1264 01:07:56,190 --> 01:07:59,020 However, question is, what happens if two hawks meet? 1265 01:07:59,020 --> 01:08:02,370 And that's a situation where they really go at it. 1266 01:08:02,370 --> 01:08:06,220 And what you might say is, then, there's a 50-50 shot, 1267 01:08:06,220 --> 01:08:10,720 that each individual has a 50% chance of getting the benefit, 1268 01:08:10,720 --> 01:08:13,222 but then also has to pay this cost. 1269 01:08:13,222 --> 01:08:14,680 And then you can decide whether you 1270 01:08:14,680 --> 01:08:17,840 want to have-- so I think it's normally parameterized 1271 01:08:17,840 --> 01:08:20,841 to b minus c over 2. 1272 01:08:20,841 --> 01:08:23,090 The idea here is that you have a 50% chance of getting 1273 01:08:23,090 --> 01:08:26,870 the benefit, but you have a 50% chance of getting beaten up. 1274 01:08:26,870 --> 01:08:28,350 OK. 1275 01:08:28,350 --> 01:08:29,359 Very simple model. 1276 01:08:29,359 --> 01:08:32,090 Now the way that we normally parameterize 1277 01:08:32,090 --> 01:08:39,837 this is that-- let's say that we want the hawk-- wait, 1278 01:08:39,837 --> 01:08:41,600 hold on a second. 1279 01:08:41,600 --> 01:08:44,250 Under which circumstances will the hawk strategy 1280 01:08:44,250 --> 01:08:46,324 be a Nash equilibrium? 1281 01:08:52,228 --> 01:08:53,720 AUDIENCE: If b is larger than c. 1282 01:08:53,720 --> 01:08:53,990 PROFESSOR: That's right. 1283 01:08:53,990 --> 01:08:55,899 If b is larger than c, then the hawk 1284 01:08:55,899 --> 01:08:58,229 would be a Nash equilibrium, which is actually what 1285 01:08:58,229 --> 01:08:59,460 we don't want in this game. 1286 01:08:59,460 --> 01:09:02,540 So what we typically assume is that the benefit 1287 01:09:02,540 --> 01:09:04,720 is less than the cost. 1288 01:09:04,720 --> 01:09:08,240 So in that sense, it's somehow not worth fighting for. 1289 01:09:08,240 --> 01:09:13,319 And that means that neither strategy is a Nash equilibrium. 1290 01:09:13,319 --> 01:09:20,747 So neither the hawk nor the dove is a Nash equilibrium. 1291 01:09:20,747 --> 01:09:24,880 And indeed you can calculate what the Nash equilibrium is, 1292 01:09:24,880 --> 01:09:30,350 and the Nash equilibrium is to follow a mixed strategy 1293 01:09:30,350 --> 01:09:34,990 with some probability p star that is, I think, b over c. 1294 01:09:38,330 --> 01:09:40,847 So if there's a bigger benefit, p star 1295 01:09:40,847 --> 01:09:42,930 is the probability of following the hawk strategy. 1296 01:09:53,859 --> 01:09:57,920 But as the costs grow, then this Nash equilibrium probability 1297 01:09:57,920 --> 01:09:59,750 decreases. 1298 01:09:59,750 --> 01:10:04,390 So this is a case where it's a mixed strategy that 1299 01:10:04,390 --> 01:10:05,680 is this Nash equilibrium. 1300 01:10:09,060 --> 01:10:11,290 And so this is a situation that displays 1301 01:10:11,290 --> 01:10:14,644 what we call negative frequency dependent selection. 1302 01:10:14,644 --> 01:10:16,560 You can see that the rare strategies do better 1303 01:10:16,560 --> 01:10:18,164 than common strategies. 1304 01:10:18,164 --> 01:10:19,580 The key way to think about this is 1305 01:10:19,580 --> 01:10:21,630 that if you have a population of doves, 1306 01:10:21,630 --> 01:10:24,050 then a rare hawk does better. 1307 01:10:24,050 --> 01:10:25,930 However, if you have a population of hawks, 1308 01:10:25,930 --> 01:10:29,150 then it's a rare dove that does better. 1309 01:10:29,150 --> 01:10:31,940 And it's that kind of negative frequency of dependence 1310 01:10:31,940 --> 01:10:35,540 that pushes the population towards some equilibrium 1311 01:10:35,540 --> 01:10:38,610 between the phenotypes, or it could 1312 01:10:38,610 --> 01:10:42,200 be equilibrium between genotypes as well, because you could have 1313 01:10:42,200 --> 01:10:47,100 pure strategists of a hawk and a dove, so different genotypes 1314 01:10:47,100 --> 01:10:48,760 that just are different. 1315 01:10:48,760 --> 01:10:51,720 And the differences in fitness will naturally 1316 01:10:51,720 --> 01:10:55,260 lead to evolution to an equilibrium hawk 1317 01:10:55,260 --> 01:10:56,840 kind of frequency in the population 1318 01:10:56,840 --> 01:10:58,580 that is equal to this. 1319 01:10:58,580 --> 01:11:01,860 So this tells us either about the equilibrium frequency 1320 01:11:01,860 --> 01:11:04,210 of the genotype of hawk in the population. 1321 01:11:04,210 --> 01:11:06,350 Or it could be that it's a single genotype that 1322 01:11:06,350 --> 01:11:10,105 implements these two strategies with some frequency that 1323 01:11:10,105 --> 01:11:10,810 is b over c. 1324 01:11:19,200 --> 01:11:23,232 So I'd say that this is a simple idea from the context of game 1325 01:11:23,232 --> 01:11:25,190 theory, and the question is, what circumstances 1326 01:11:25,190 --> 01:11:28,315 might this actually arise in kind of real populations. 1327 01:11:32,215 --> 01:11:33,590 Oh, before I say anything else, I 1328 01:11:33,590 --> 01:11:34,964 do want to stress that all of you 1329 01:11:34,964 --> 01:11:37,670 have played mixed strategies before, because in the Rock 1330 01:11:37,670 --> 01:11:40,086 Paper Scissors game, which we're going to be talking about 1331 01:11:40,086 --> 01:11:45,580 more in the coming week or so-- you guys all know this, 1332 01:11:45,580 --> 01:11:48,820 that rock beats-- not paper-- but rock beats scissors, 1333 01:11:48,820 --> 01:11:50,930 scissors beats paper, paper beats rock-- 1334 01:11:50,930 --> 01:11:55,360 so the Nash equilibrium of that game is to do 1/3 1/3, 1/3, 1335 01:11:55,360 --> 01:11:57,360 because if everybody's following that strategy, 1336 01:11:57,360 --> 01:11:59,330 nobody has the incentive to change strategies. 1337 01:11:59,330 --> 01:12:01,300 Definition of the Nash equilibrium. 1338 01:12:01,300 --> 01:12:05,080 Is that the optimal strategy against any opponent? 1339 01:12:05,080 --> 01:12:06,330 No. 1340 01:12:06,330 --> 01:12:08,100 And this is very important to point out, 1341 01:12:08,100 --> 01:12:10,870 because when you think about these Nash equilibrium 1342 01:12:10,870 --> 01:12:12,750 and everything in abstract terms, 1343 01:12:12,750 --> 01:12:14,640 it's easy to get confused about-- 1344 01:12:14,640 --> 01:12:17,016 when we say that it's-- sometimes we say optimal, 1345 01:12:17,016 --> 01:12:18,390 or it's the solution to the game, 1346 01:12:18,390 --> 01:12:20,830 that doesn't mean that it's the best response to any given 1347 01:12:20,830 --> 01:12:21,690 strategy. 1348 01:12:21,690 --> 01:12:23,064 And you know that, because if you 1349 01:12:23,064 --> 01:12:26,012 know that your little brother is going to play rock, then 1350 01:12:26,012 --> 01:12:27,720 you know-- well, depending on whether you 1351 01:12:27,720 --> 01:12:30,220 want to make him happy or not, you decide what to do, right? 1352 01:12:30,220 --> 01:12:32,970 So it's clear that this 1/3, 1/3, 1/3 thing 1353 01:12:32,970 --> 01:12:35,197 is, it can be the solution to the game of the Nash 1354 01:12:35,197 --> 01:12:37,530 equilibrium without being the best response to any given 1355 01:12:37,530 --> 01:12:38,029 strategy. 1356 01:12:41,384 --> 01:12:43,300 So one of things that we're doing in my group, 1357 01:12:43,300 --> 01:12:47,370 actually, is we're arguing that in the case of yeasts that 1358 01:12:47,370 --> 01:12:50,720 are exposed to mixed sugar environments-- 1359 01:12:50,720 --> 01:12:53,850 so yeast in environments that contain a little bit of glucose 1360 01:12:53,850 --> 01:12:56,415 and a little bit of galactose-- what we and actually 1361 01:12:56,415 --> 01:13:00,420 other people have found is that there's a bimodal response. 1362 01:13:00,420 --> 01:13:08,200 So if you look at the number of cells as a function of the GAL 1363 01:13:08,200 --> 01:13:10,472 operon expression, so GAL is the expression 1364 01:13:10,472 --> 01:13:12,930 of the genes required to break down galactose, what you see 1365 01:13:12,930 --> 01:13:17,240 is that some of the cells activate the GAL genes highly, 1366 01:13:17,240 --> 01:13:19,620 some don't activate it at all. 1367 01:13:19,620 --> 01:13:21,420 So the question is, what might be 1368 01:13:21,420 --> 01:13:23,810 an explanation for this phenotypic heterogeneity 1369 01:13:23,810 --> 01:13:26,190 observed in a clonal population? 1370 01:13:26,190 --> 01:13:27,960 And one way to think about this is 1371 01:13:27,960 --> 01:13:29,610 that it could be the implementation 1372 01:13:29,610 --> 01:13:31,420 of some sort of foraging game. 1373 01:13:31,420 --> 01:13:34,530 So you can imagine a population of animals that want 1374 01:13:34,530 --> 01:13:36,320 to go and eat some berries. 1375 01:13:36,320 --> 01:13:38,050 So there are two different food sources-- 1376 01:13:38,050 --> 01:13:41,267 say there are blueberries and there are red berries. 1377 01:13:41,267 --> 01:13:43,350 You can imagine this is the kind of situation that 1378 01:13:43,350 --> 01:13:45,183 would lead to negative frequency dependence, 1379 01:13:45,183 --> 01:13:47,464 because if everybody else goes and eats-- 1380 01:13:47,464 --> 01:13:49,880 I guess this was the blueberry-- so if everybody else goes 1381 01:13:49,880 --> 01:13:51,360 and eats the blueberries, then it's better for you 1382 01:13:51,360 --> 01:13:52,750 to eat the red berry because you don't have 1383 01:13:52,750 --> 01:13:54,830 to share with anyone, whereas if everyone else goes and eats 1384 01:13:54,830 --> 01:13:56,650 the red berries, then it's better for you 1385 01:13:56,650 --> 01:13:58,460 to eat the blueberries. 1386 01:13:58,460 --> 01:14:01,050 So that's the kind of situation that would naturally 1387 01:14:01,050 --> 01:14:03,860 lead to the so-called negative frequency dependent selection, 1388 01:14:03,860 --> 01:14:06,644 in which coexistence of the phenotypes 1389 01:14:06,644 --> 01:14:08,185 is somehow the equilibrium condition, 1390 01:14:08,185 --> 01:14:11,220 and that can be implemented either by different genotypes 1391 01:14:11,220 --> 01:14:13,830 or it can be implemented by a single genotype that 1392 01:14:13,830 --> 01:14:17,050 follows both phenotypes. 1393 01:14:17,050 --> 01:14:20,772 Incidentally, is this equilibrium-- 1394 01:14:20,772 --> 01:14:22,730 does it maximize the fitness of the population? 1395 01:14:34,260 --> 01:14:34,760 No. 1396 01:14:34,760 --> 01:14:36,210 So what is it that would maximize 1397 01:14:36,210 --> 01:14:37,460 the fitness of the population? 1398 01:14:39,720 --> 01:14:42,129 If everybody were a dove, because-- and you can 1399 01:14:42,129 --> 01:14:43,670 you go ahead and do this calculation, 1400 01:14:43,670 --> 01:14:45,420 find out what the fitness of everybody is, 1401 01:14:45,420 --> 01:14:49,300 but what you'll see is that p star does not maximize fitness, 1402 01:14:49,300 --> 01:14:52,800 ford not maximize payoff. 1403 01:14:52,800 --> 01:14:56,799 So in these games, the equilibrium is not necessarily 1404 01:14:56,799 --> 01:14:57,590 fitness maximizing. 1405 01:15:09,620 --> 01:15:12,280 So so far what we've done is we've 1406 01:15:12,280 --> 01:15:15,510 given two examples of different evolutionary drivers 1407 01:15:15,510 --> 01:15:17,550 for phenotypic heterogeneity in a population. 1408 01:15:17,550 --> 01:15:18,924 And I just want to highlight what 1409 01:15:18,924 --> 01:15:24,360 I think is maybe the last big one, which would be altruism. 1410 01:15:24,360 --> 01:15:27,480 And sometimes people call it division of labor, 1411 01:15:27,480 --> 01:15:29,890 or sacrifice or so. 1412 01:15:29,890 --> 01:15:33,402 And a nice example of this is colicin production. 1413 01:15:40,790 --> 01:15:44,600 The idea is that in many bacteria, 1414 01:15:44,600 --> 01:15:47,765 they express these colicin, which are toxins. 1415 01:15:47,765 --> 01:15:49,940 And we're going to read more about this later. 1416 01:15:49,940 --> 01:15:51,555 So colicin is a toxin. 1417 01:15:55,020 --> 01:15:57,430 And the perhaps surprising thing is 1418 01:15:57,430 --> 01:16:03,497 that in many gram-negative bacteria, like E. coli, when 1419 01:16:03,497 --> 01:16:05,580 they make this colicin, the only way it's released 1420 01:16:05,580 --> 01:16:06,440 is cell lysis. 1421 01:16:16,770 --> 01:16:21,200 Now the question is, why would I possibly do this? 1422 01:16:21,200 --> 01:16:24,150 And the answer is, perhaps, that if there's 1423 01:16:24,150 --> 01:16:27,460 a plasmid that encodes the genes to make this toxin, 1424 01:16:27,460 --> 01:16:30,940 it also encodes an immunity protein. 1425 01:16:30,940 --> 01:16:35,090 What that means is that if one cell lyses-- 1426 01:16:35,090 --> 01:16:37,280 so it bursts open and releases the colicin-- 1427 01:16:37,280 --> 01:16:40,010 it will inhibit other bacteria, but not 1428 01:16:40,010 --> 01:16:43,880 the bacteria that are the clone mates that are related, 1429 01:16:43,880 --> 01:16:47,065 so not the bacteria that also carry that plasmid. 1430 01:16:47,065 --> 01:16:48,940 So it's important to note there that now when 1431 01:16:48,940 --> 01:16:51,760 we talk about kin selection or relatedness, 1432 01:16:51,760 --> 01:16:54,960 we're talking about the other cells that carry that gene. 1433 01:16:54,960 --> 01:17:00,990 So what's relevant is that the plasma that encodes this toxin 1434 01:17:00,990 --> 01:17:05,580 will inhibit other bacteria that do not encode the toxin. 1435 01:17:05,580 --> 01:17:08,560 Now you can see that this kind of lysis behavior 1436 01:17:08,560 --> 01:17:10,589 has to be stochastic, so there has 1437 01:17:10,589 --> 01:17:11,880 to be phenotypic heterogeneity. 1438 01:17:11,880 --> 01:17:12,546 And why is that? 1439 01:17:17,280 --> 01:17:18,690 AUDIENCE: Everyone always lyses. 1440 01:17:18,690 --> 01:17:19,690 PROFESSOR: That's right. 1441 01:17:19,690 --> 01:17:21,610 If this plasmid always makes everybody lyse, 1442 01:17:21,610 --> 01:17:24,240 then it's not going to get very far, right? 1443 01:17:24,240 --> 01:17:26,620 But you can kind of imagine situations 1444 01:17:26,620 --> 01:17:29,270 where, if you just 1% of the population 1445 01:17:29,270 --> 01:17:31,700 lyses but then inhibits the cells that 1446 01:17:31,700 --> 01:17:33,832 don't care that plasmid, then the plasmid 1447 01:17:33,832 --> 01:17:35,040 can spread in the population. 1448 01:17:38,810 --> 01:17:40,740 So this has to be stochastic. 1449 01:17:51,030 --> 01:17:53,280 So I think that this is another general explanation 1450 01:17:53,280 --> 01:17:55,238 for why there might be phenotypic heterogeneity 1451 01:17:55,238 --> 01:17:56,570 in the population. 1452 01:17:56,570 --> 01:17:59,560 All of these three explanations are both conceptually different 1453 01:17:59,560 --> 01:18:01,460 and they're experimentally different. 1454 01:18:01,460 --> 01:18:05,070 They make different falsifiable predictions 1455 01:18:05,070 --> 01:18:08,690 about what's driving that phenotypic heterogeneity. 1456 01:18:08,690 --> 01:18:10,740 In this field, I think often when people 1457 01:18:10,740 --> 01:18:12,660 see phenotypic heterogeneity, they immediately 1458 01:18:12,660 --> 01:18:16,080 assume that it's bet hedging, but these other two 1459 01:18:16,080 --> 01:18:18,660 explanations, I think, are just as general 1460 01:18:18,660 --> 01:18:20,416 and may be just as common. 1461 01:18:20,416 --> 01:18:22,290 So we really have to go and make measurements 1462 01:18:22,290 --> 01:18:26,280 to try to elucidate what's going on in each case.