Evolvability

Top

00:00:07.03 Okay, well, my name is Marc Kirschner.
00:00:09.20 I'm from Harvard University.
00:00:11.23 And I want to talk in this lecture about evolution
00:00:15.17 from a more modern perspective
00:00:18.10 of cell and developmental biology
00:00:20.15 and molecular biology.
00:00:22.17 And I really want to explain this word evolvability,
00:00:25.03 which is hard to pronounce.
00:00:27.15 They say if you say it three times quickly,
00:00:29.16 everybody chokes on the word.
00:00:32.23 And some people think it's maybe something
00:00:35.17 that contradicts Darwin's theory of evolution...
00:00:38.21 which of course it doesn't contradict Darwin's theory of evolution.
00:00:43.13 And it augments it, really.
00:00:46.02 And some people think... in the evolutionary community, at one time,
00:00:50.27 they thought this word might have some sense of orthogenesis,
00:00:55.02 that is, some predestination of evolution
00:00:59.22 to a certain end, like human beings.
00:01:02.04 And of course, it doesn't have that as well.
00:01:04.24 So, we're gonna define this word by example
00:01:08.10 and show why it's powerful and important
00:01:12.06 to understand in evolutionary theory.
00:01:15.04 Now, we can all agree that evolution
00:01:18.04 is successful and creative.
00:01:20.02 It's been so creative and so successful
00:01:24.19 that people doubted that it really
00:01:31.01 could be explained by natural principles,
00:01:33.04 such as those of Darwin.
00:01:35.07 It's still hard to imagine
00:01:37.28 that out of some primeval ooze
00:01:41.04 all these various beautiful life forms
00:01:43.25 appeared on this planet.
00:01:45.26 And I think what we'll talk about today
00:01:50.06 may be something which helps explain
00:01:52.26 the ease by which these things happened.
00:01:54.23 So, it doesn't explain how each one happened,
00:01:57.01 but it explains some of the underlying processes
00:01:59.02 that are not too obvious,
00:02:01.18 unless you delve deep... dig deeply
00:02:05.03 into the molecular and cellular processes underneath them.
00:02:09.26 So, let's start with Darwin's big idea.
00:02:12.26 He had two really major concepts in his idea.
00:02:17.12 He wrote them down in his notebook
00:02:20.27 when he came back from the voyage of the Beagle,
00:02:23.18 and that was published.
00:02:26.07 And that is that all life on Earth comes from a common origin.
00:02:29.23 So, you see that sort of tree at the bottom
00:02:32.08 with its common origin on the left,
00:02:34.19 and the various branches all coming from the same source.
00:02:38.28 Remember, he said that before we knew
00:02:41.27 as much as we do
00:02:44.17 about the underlying molecular similarity among organisms.
00:02:48.11 And he also believed... the second big idea
00:02:51.21 is it proceeds by a process of heritable change and selection.
00:02:55.15 And this selection is on the organism itself,
00:02:58.21 on what we call the phenotype,
00:03:01.12 and that includes its anatomy, its physiology,
00:03:04.23 and its behavior.
00:03:07.12 And again, if you look at the bottom on the right,
00:03:10.08 you see the anatomy of the finches that he found on...
00:03:13.20 the various species of finches
00:03:16.07 that he found on the islands of the Galapagos,
00:03:19.08 which we now believe originated
00:03:22.25 from a single founding species
00:03:25.12 that under the pressure of selection,
00:03:27.26 which was mostly the nature of the food
00:03:30.02 and the temperature,
00:03:32.02 developed various different shapes
00:03:34.22 appropriate to the kind of food that was available to them.
00:03:37.24 And that very much influenced him as well.
00:03:41.14 Now, here's the problem.
00:03:43.08 We understand selection.
00:03:46.18 That's something that Darwin really based
00:03:49.27 a lot on animal husbandry and plant selection.
00:03:53.10 We can see that in terms of human intervention.
00:03:57.18 And... but where does the novelty come from
00:04:02.15 on which selection can act?
00:04:05.14 And he believed that there was
00:04:10.27 random variation in the phenotype
00:04:13.07 which was heredit... which was inherited.
00:04:15.17 Of course, he didn't know at that time
00:04:17.18 how it was inherited.
00:04:20.02 And he made certain assumptions
00:04:22.10 about the nature of that variation.
00:04:25.14 He assumed that variation was plentiful,
00:04:28.04 so that selection had a lot to act on.
00:04:31.25 So, all these various forms that we see
00:04:34.23 are a result of selection going
00:04:39.16 in many, many different directions.
00:04:42.05 He believed that variation was like clay,
00:04:44.27 capable of being molded in every direction,
00:04:47.11 so that selection could choose any of all possible
00:04:51.17 variant phenotypes.
00:04:53.25 And he also believed that each step of variation
00:04:56.28 was small, even imperceptible.
00:05:00.11 Now, those were his assumptions.
00:05:05.29 He didn't really have evidence to support them
00:05:09.13 except these he felt were necessary
00:05:12.01 for the theory to be successful.
00:05:15.10 But there's a serious problem.
00:05:17.13 How does a complex structure like the eye
00:05:20.02 emerge from small evolutionary steps,
00:05:23.13 if it’s only functional
00:05:27.00 when all parts are assembled?
00:05:28.23 So, here we see a drawing of the eye,
00:05:30.20 and we see the lens, which is necessary,
00:05:32.22 we see the retina, which is necessary,
00:05:35.09 we see the optic nerve,
00:05:38.05 and of course in the brain we'll need all the parts of the brain,
00:05:40.14 which are able to interpret the images.
00:05:43.00 So, how could this all happen
00:05:46.03 when it only would be successful
00:05:48.19 after all these parts are brought together?
00:05:50.29 What allowed them to continue to be selected
00:05:53.07 without any actually functional value?
00:05:56.14 And this bothered Darwin a great deal.
00:05:58.20 He made very good arguments
00:06:01.00 about how this might take place,
00:06:03.10 but it's always plagued the theory of evolution,
00:06:05.13 for a long time in the eyes of skeptics.
00:06:08.06 For example, how would a bird develop its wing,
00:06:11.10 if only after it could fly
00:06:15.04 it could be selected to be a better flyer?
00:06:17.09 So, people have said,
00:06:19.29 well, Darwin's theory is a good theory for improvement,
00:06:23.04 but not a very good theory for innovation.
00:06:27.03 And we'll have to come to that.
00:06:29.15 What level of innovation
00:06:32.05 is being used to generate
00:06:35.06 these parallel processes?
00:06:37.15 And how can they be selected together?
00:06:40.09 Well, I think we are beginning to see that
00:06:43.08 novelty is not what it used to be.
00:06:45.11 There's more knowledge about how novelty
00:06:47.24 is generated then was extant in Darwin's day,
00:06:50.26 when he really didn't even have
00:06:53.06 at his disposal a good theory of the cell.
00:06:57.15 First, we've seen that steps can be large.
00:07:00.22 A few genetic changes can significantly change the phenotype.
00:07:05.00 Next, we can see that even big changes
00:07:07.09 are rarely lethal.
00:07:09.26 You can have a fly that generates
00:07:12.14 from two to four wings,
00:07:14.28 and it still lives.
00:07:16.22 All at once, and in big steps,
00:07:18.22 very quickly, non-lethal.
00:07:21.19 And I would argue
00:07:24.03 -- and this has something you could... you could question,
00:07:27.13 but we'll see it come back later --
00:07:29.18 that novelty isn't just random;
00:07:32.21 it's biased toward useful change.
00:07:34.26 Now, how can somebody be biased for useful change?
00:07:37.13 If you're making something novel,
00:07:39.08 how would you know it would be useful?
00:07:41.19 But it may be a modification of something
00:07:43.20 which has been useful in some other form.
00:07:46.06 For example, if we look at the vertebrate appendage
00:07:49.04 -- vertebrate, let's say, arms and hands --
00:07:52.05 those are really all modified forms
00:07:55.16 of something which have been useful in other forms.
00:07:58.04 They may be flippers on the whale.
00:08:00.10 They may be wings on a bird.
00:08:02.13 But the basic idea of capturing space is similar.
00:08:08.07 And so, there's a lot of utility involved
00:08:10.12 in the kinds of changes, at least,
00:08:12.17 that have been selected -- common utility.
00:08:15.13 So, what I want to really be talking about is this question,
00:08:19.12 as I say, of evolvability -- this unpronounceable word --
00:08:22.25 as a process of generating the right kind of novelty.
00:08:26.05 And it's the capacity of an organism
00:08:30.12 to generate novelty
00:08:32.23 that tends to be non-lethal,
00:08:35.07 because any novelty that may be useful
00:08:38.05 to the individual in one generation, which is not heritable
00:08:41.01 -- is not passed on to the next generation --
00:08:44.14 does not contribute to evolution.
00:08:47.06 And if the animal just dies because of the novelty,
00:08:49.16 it's not useful either.
00:08:52.10 And it's also a kind of novelty
00:08:54.10 that's likely to be useful,
00:08:56.08 because there are so many directions in which novelty can go
00:08:59.25 that if you bias it in some direction
00:09:01.28 that may be more useful,
00:09:03.29 you'll have a much larger likelihood
00:09:06.04 of generating something which will be useful,
00:09:08.10 and selected for in evolution.
00:09:10.23 So, here are four general types of evolvability
00:09:15.06 that I want to talk about.
00:09:17.02 One is regulatory change,
00:09:20.28 which is changing the time, the place,
00:09:23.18 the circumstances under which something happens.
00:09:27.14 Another is exploratory behavior,
00:09:29.17 which I'll define more,
00:09:31.09 but it has to do with a process
00:09:33.29 which on its own will generate many different states.
00:09:37.17 A third is compartmentation,
00:09:40.13 which is a real problem...
00:09:42.18 is how you make modifications in one part of the anatomy of the organism
00:09:47.20 without making it in other parts of the organism.
00:09:50.11 So, you may want to... to put,
00:09:53.03 you know, hair in certain parts of the anatomy
00:09:55.13 and not in other parts of the anatomy.
00:09:57.18 These are the kinds of things that have to be worked out.
00:09:59.28 How is that done on the molecular level?
00:10:02.12 And the fourth is something called weak linkage,
00:10:04.22 which is the ease in which
00:10:08.20 one can make new connections
00:10:11.08 without changing very much in evolution.
00:10:14.07 So, let's start with regulatory change.
00:10:16.25 So, as I say, it's a change in time
00:10:19.22 when something happens, time and place,
00:10:23.04 time and amount
00:10:25.02 -- the degree to which something happens
00:10:26.25 -- or circumstances.
00:10:29.18 And at the level of genes, we're talking about how a gene is used.
00:10:31.28 It's not a change in the structure of the gene product.
00:10:35.05 So, here's a perfectly good example
00:10:37.24 of how changing
00:10:41.18 the place in which genes operate
00:10:44.26 and changing the amount
00:10:47.15 can give you quite a bit of variety,
00:10:49.20 which has been very useful to man,
00:10:51.24 and that is the husbandry and selection
00:10:55.18 that human beings had on the wild mustard plant.
00:11:00.10 And if they selected for variations in the stem,
00:11:04.11 they got kohlrabi.
00:11:06.09 If they selected for lateral buds,
00:11:08.14 they got Brussels sprouts.
00:11:10.17 If they selected for terminal buds,
00:11:12.11 they got cabbage.
00:11:14.14 If they selected for flower clusters,
00:11:16.11 they got cauliflower.
00:11:18.06 If they selected for stems and flowers,
00:11:20.06 they got broccoli.
00:11:22.00 And if they selected for leaves,
00:11:23.19 they got kale.
00:11:25.20 So, all this -- just time and manner and place --
00:11:29.17 without really changing the basic anatomy of the plant,
00:11:32.20 gave many, many different forms of the plant,
00:11:35.28 which we found very useful.
00:11:38.08 Let's talk about the case of invention of flight.
00:11:42.11 In vertebrates, as...
00:11:45.15 I'm not talking about insects, of course...
00:11:47.11 but in vertebrates, flight was invented independently three times.
00:11:51.24 Once, on the right, is the bat,
00:11:55.04 which is the mammalian form of wings and flight.
00:11:59.21 The middle is a bird, a pelican,
00:12:02.12 which is the bird form of wings and flight.
00:12:04.23 And on the left is a pterodactyl.
00:12:07.02 I'm very proud of the pterodactyl picture.
00:12:09.13 It's very hard to get pictures,
00:12:12.02 these days, of flying pterodactyls.
00:12:14.13 But that turns out to be a separate invention
00:12:17.14 of flight.
00:12:20.07 How did evolution evolve flight
00:12:23.29 these three separate ways?
00:12:25.25 Was it all the same thing?
00:12:27.19 Was there something magical and new?
00:12:29.21 And when you look at just the skeletons alone,
00:12:31.17 which of course are well preserved in the fossil record,
00:12:34.03 so we have a tremendous amount of data on this,
00:12:37.14 you see that...
00:12:40.10 you know, that much of the animals is the same.
00:12:42.23 But the changes really occur,
00:12:44.15 as you would expect, in the wings.
00:12:47.05 But, how did they change?
00:12:50.07 There were really no new genes
00:12:53.02 that were needed to structurally support flight.
00:12:55.05 It was really about the amount and the timing
00:12:58.12 and the location in which bone growth occurred.
00:13:02.18 So, in the pterodactyl,
00:13:05.03 you extended the bone of the fourth digit.
00:13:08.18 In the bird,
00:13:10.23 you extended the forearm and wrist bones.
00:13:12.28 And in the bat, which... over here,
00:13:15.19 you have the forearm, which is the ulna bone,
00:13:20.29 and the third, fourth, and fifth digits,
00:13:23.14 which are extended.
00:13:26.16 And it's really just about when you terminate bone growth
00:13:30.02 in these animals,
00:13:32.06 and you can generate a lot of other things.
00:13:34.04 Now, you could argue,
00:13:36.00 well, you need more than just bones.
00:13:38.09 You need skin to cover those bones.
00:13:40.27 But if anybody's ever seen a woman
00:13:44.21 who is pregnant,
00:13:46.17 you realize getting more skin to cover the woman's belly
00:13:48.14 is not a problem,
00:13:50.06 not to mention all the people who are obese, or they...
00:13:54.16 or people who have burns that heal.
00:13:57.01 So, skin is something which,
00:13:59.13 you know, you wouldn't probably even need to change anything about skin
00:14:01.28 to have it covering the bones and make...
00:14:04.01 and make a wing.
00:14:06.03 The muscles... the same muscle groups are involved,
00:14:08.16 but they're bigger, of course,
00:14:10.19 if they're flying than they would be.
00:14:12.19 But that's something, at least,
00:14:14.27 you could do over time by selection on the...
00:14:17.09 on the size,
00:14:19.12 that is, the amount of muscles of certain muscles.
00:14:22.13 So, it's the change in regulation of length,
00:14:24.13 not how bones are made
00:14:26.16 or the invention of new bones.
00:14:29.13 Now, at... the question about a change is really...
00:14:36.19 the 20th century taught us that things
00:14:40.06 are more similar than we thought,
00:14:42.18 which is, in some sense,
00:14:45.00 reassuring from Darwin's first view
00:14:47.22 that we all originated from some common origin,
00:14:50.20 but makes it more a question, more puzzling,
00:14:54.21 about where change actually occurs.
00:14:57.01 If everything is the same, then we don't have to...
00:14:59.19 we really have to find a place to explain the differences.
00:15:03.10 Now, the similarities are quite striking.
00:15:06.01 The human genome has about 21,000 genes.
00:15:08.21 That's only one and a half times the number of fruit fly genes,
00:15:11.24 and only six times the number of a single gut bacterium,
00:15:15.02 E. coli.
00:15:17.21 And we share 15% of our genes with E. coli,
00:15:24.02 25% of our genes with yeast,
00:15:27.11 50% with a fruit fly,
00:15:29.18 and 70% with a frog.
00:15:31.23 So, there's a fair amount of conservation
00:15:35.09 that's going on.
00:15:37.25 And does this conservation make it more difficult to change,
00:15:40.24 which one may argue it does,
00:15:43.29 or is it even possible
00:15:47.02 that it's a certain kind of conservation
00:15:50.06 that actually facilitates change?
00:15:52.07 And we'll explore that.
00:15:53.28 Okay.
00:15:56.15 So, let's imagine some of the problems
00:15:59.11 with making a limb.
00:16:01.05 We not only have to have the bones;
00:16:02.27 we have to have the muscles there to provide power;
00:16:04.18 we have to control those muscles with nerves;
00:16:06.20 we have to provide a blood supply
00:16:10.04 for food and to remove wastes;
00:16:12.17 and everything has to be placed properly for the wing to work.
00:16:16.16 That's still a tall order.
00:16:19.05 Making the bones is hard enough,
00:16:22.06 but getting all these things to happen...
00:16:24.19 and they have to happen in some coordination with the...
00:16:28.22 with the development of the bones for us to have a functional wing.
00:16:33.17 And to help explain that,
00:16:36.01 we need to talk about something which is quintessentially biological,
00:16:38.15 which we call exploratory behavior.
00:16:42.08 And I'll give examples of it,
00:16:44.27 and there are many more examples one could choose,
00:16:48.23 but I'll give them with the idea of trying
00:16:51.24 to explain some of the changes in forming the wing.
00:16:54.06 So, let me just define it.
00:16:56.10 It's a... it... exploratory behavior
00:16:58.17 is a little like evolution,
00:17:00.24 except it's not genetic.
00:17:02.27 It's a non-genetic process of random variation
00:17:05.08 followed by selection for a functional outcome.
00:17:07.09 And we're not talking about this
00:17:09.17 in the course of evolution,
00:17:11.09 but in the course of the life of the animal.
00:17:13.06 In the process of exploration,
00:17:15.06 as in any exploration,
00:17:16.29 many futile paths are taken
00:17:19.27 to maximize the chance of finding a functional one.
00:17:22.07 So, here's a wonderful example.
00:17:24.17 It's not of an anatomy,
00:17:26.15 but it's a behavioral example of exploratory behavior.
00:17:30.10 And it's a question of how ants forage for food.
00:17:33.29 Now, ants don't talk to each other, particularly.
00:17:38.25 They can... they can communicate some, of course,
00:17:42.18 chemically.
00:17:44.29 And they don't see very well.
00:17:46.20 They can kind of see,
00:17:48.23 you know, light and dark.
00:17:50.14 Maybe there's a little bit of imaging,
00:17:52.09 but not much at all.
00:17:53.27 So, there's no chance, emerging from the ant...
00:17:56.11 from the anthill,
00:17:58.21 that they're gonna be able to see the food in the environment.
00:18:01.02 So, they explore randomly.
00:18:03.26 And what they do is they...
00:18:07.21 as they explore, they leave a trail behind them of an...
00:18:11.03 of an odorant, a pheromone.
00:18:14.18 It's a little like Hansel and Gretel.
00:18:16.21 They leave a little... they leave cracker chips.
00:18:20.15 And then they... after they've gone for a certain time
00:18:23.15 and they haven't found any food,
00:18:25.16 they turn around and follow the...
00:18:28.19 their path by smell back to the anthill --
00:18:33.00 otherwise, they wouldn't be able to find the anthill.
00:18:35.22 And another ant might do the same.
00:18:37.13 Mission failed. Went off a certain distance.
00:18:39.15 Comes around and comes back.
00:18:41.14 But one ant may have found a lusc...
00:18:42.25 a delicious dead cockroach out there
00:18:47.05 in the environment.
00:18:49.02 And when it goes back,
00:18:51.17 it secretes more pheromone and reinforces the trail.
00:18:55.05 And ants tend to follow trails.
00:18:57.19 And they...
00:19:01.02 and so the ones that have come up with this...
00:19:04.07 it's a volatile trail, so it doesn't last very long,
00:19:06.17 but another ant coming out
00:19:09.00 may smell that trail and follow it,
00:19:10.28 find that same delicious cockroach out there,
00:19:13.01 reinforce the trail further.
00:19:14.25 And over time, all the ants will be following
00:19:17.09 the major trail to the cockroach
00:19:19.15 and not the futile trails to nothing.
00:19:22.19 And that's... that's a way in which
00:19:26.17 you get a structured outcome
00:19:29.07 without any inherent structure,
00:19:31.14 just by lots of variation and reinforcement.
00:19:34.15 Now, there’s a problem with this method
00:19:37.04 in that you end up finding the first good food,
00:19:39.16 not the best good food,
00:19:41.12 and so this is something,
00:19:43.08 if there are any graduate students or students generally in the audience...
00:19:45.27 that there's a role for lost ants.
00:19:47.28 That is, for reasons we don't understand,
00:19:50.05 ants who are following a big trail
00:19:52.15 occasionally wander off the trail to some distance
00:19:56.12 and so that's... that's telling you essentially that,
00:19:59.24 yes, you could exploit what your professor
00:20:01.20 has told you is a... is a goldmine of results,
00:20:04.03 but if your mind tells you
00:20:07.03 you might want explore something else,
00:20:08.22 it may be worth exploring it.
00:20:11.00 Alright.
00:20:12.19 So, that's a kind of a non-anatomical way,
00:20:14.09 but let's go back to the limb.
00:20:17.13 Bone formation is the leading event.
00:20:20.17 But everything else is exploratory and adaptive.
00:20:25.22 It's not that it has to be told what to do;
00:20:27.19 it finds its path by variation and selection.
00:20:30.24 So, the first thing after you made this limb-like...
00:20:34.18 limb bud and have this limb forming
00:20:37.15 is you have the somite tissue,
00:20:40.10 which has the muscle precursors, myoblasts, in them,
00:20:43.04 and they migrate out in all directions.
00:20:45.11 But they find their attachment next to the bones,
00:20:48.18 so wherever the bones are,
00:20:51.12 the myoblasts end up being there just by random exploration.
00:20:54.02 And that's gonna generate the power.
00:20:56.19 The nerves also grow out randomly,
00:21:00.08 and I'll tell you...
00:21:02.04 give you an example of how that works.
00:21:04.03 And the blood...
00:21:06.24 that's also true of the food supply in the blood.
00:21:09.29 So, random exploration makes this very adaptive.
00:21:12.10 It means that it isn't so difficult to get novelty.
00:21:14.12 You don't need five things happening at once.
00:21:17.11 One thing can set the set the direction,
00:21:21.10 and the other things will just adapt to it.
00:21:24.20 So, if we look at the nerves coming out of the...
00:21:30.22 let's say the motor neurons that are coming out of the spinal cord,
00:21:33.08 they migrate out in all sorts of random directions
00:21:35.23 until one of them hits the target,
00:21:38.19 and that stabilizes that nerve.
00:21:41.11 And you then end up with the proper connection.
00:21:44.14 So, it looks kind of superficially
00:21:48.15 as if that nerve knew where to go.
00:21:50.20 If you follow the actual process itself, it's...
00:21:54.07 it's just stabilized.
00:21:56.10 In the same way,
00:21:58.16 it looks like flies fly toward flypaper,
00:22:00.18 but it's not true.
00:22:02.12 They're not attracted to fly paper; they just get stuck there.
00:22:05.13 So, that's the same principle at work.
00:22:07.09 So, the next question I want to ask is,
00:22:09.11 how can these changes be limited
00:22:12.08 to certain selections... sections of the body?
00:22:14.14 How is that achieved?
00:22:16.11 And this is by a process
00:22:18.14 that I don't think we could have anticipated.
00:22:20.17 There was no real sort of natural precursor
00:22:25.02 to thinking this way
00:22:27.24 except to really study the biology in real depth.
00:22:31.06 And this is a process called compartmentation.
00:22:34.23 Now, in the embryo...
00:22:37.05 this is how we'll talk about multicellular animals...
00:22:41.11 there's a pre-established map
00:22:43.12 -- not completely pre-established from the egg, but very early on --
00:22:47.08 that divides the embryo into spatial domains.
00:22:49.05 And each domain behaves differently
00:22:54.01 in development and the adult.
00:22:56.08 So, this domain... it's like a...
00:22:58.15 it's like the map of the world,
00:23:03.01 which has all these countries,
00:23:05.27 and so you see it on the map,
00:23:07.26 and one says it's Brazil and another says it's Paraguay
00:23:11.20 and another says it's China.
00:23:14.11 And they... they're there, and they don't really make any sense.
00:23:17.01 You know, they're not...
00:23:18.24 they don't correspond necessarily to geography or anything,
00:23:20.27 but they do divide the world up
00:23:23.22 so well that by knowing where the...
00:23:27.01 where something is in terms of that country,
00:23:29.13 you can place it on the globe.
00:23:31.07 So, there's a map that's generated early in development
00:23:34.12 that has every place in the embryo...
00:23:38.06 gives every place in the embryo an address.
00:23:40.16 So, compartments can be visualized
00:23:43.23 early in the embryo.
00:23:45.14 This is a Drosophila embryo.
00:23:47.06 And these are compartments that are...
00:23:49.20 from the ant... they go from the back to the front of the embryo,
00:23:52.23 and they're spaced from the anterior
00:23:56.26 to the posterior end of the embryo.
00:23:59.09 And the domains in the fly embryo,
00:24:02.04 which is where this was discovered,
00:24:05.10 are these anterior-posterior vertical lines.
00:24:10.01 But there are also dorsal-ventral lines,
00:24:13.19 and so there are... flies have about 100 such compartments.
00:24:18.05 Vertebrates have about 200 compartments.
00:24:20.23 These happen rather early in development.
00:24:24.00 And each department...
00:24:27.13 compartment is specified by
00:24:29.22 a set of transcription factors
00:24:32.23 and secreted signaling molecules.
00:24:35.01 So, it's not that each compartment
00:24:37.13 has a unique transcription factor;
00:24:39.12 it's the set that is different
00:24:41.19 from the next compartment.
00:24:43.23 And each compartment in the developing spine, for example,
00:24:48.00 which is our... you know, the vertebrate example,
00:24:50.12 tells bones to be different.
00:24:55.11 So, for example,
00:24:58.02 if we look at the bones in the neck -- the cervical vertebrae --
00:25:00.05 they take a different shape
00:25:02.24 than the bones in the thorax,
00:25:04.19 which have ribs, or the lumbar vertebrae,
00:25:07.12 in the lower back.
00:25:10.15 And if you were to take, early on in development,
00:25:13.18 some of the cells from the... from the...
00:25:18.18 say, the thorax,
00:25:21.01 and graft them into the cervical,
00:25:23.13 they would make ribs in your neck.
00:25:25.06 So, the cells are told -- early in development --
00:25:29.04 that if you develop in this area,
00:25:31.04 you will have a certain kind of shape.
00:25:33.01 You will respond to the same signals differently.
00:25:37.20 And it's this compartmentation
00:25:40.02 which is preserved through 500 million years of evolution...
00:25:45.02 what happens underneath may change,
00:25:49.19 but the identity of those compartments remains.
00:25:53.20 So, you know, it may be that, for example,
00:25:56.00 you'll have fewer ribs
00:26:00.24 in that thoracic compartment.
00:26:02.27 That may be the rules for a different species.
00:26:05.16 But what a cell does in that compartment
00:26:08.03 is specified by this early map.
00:26:11.09 And so, that's what distinguishes
00:26:14.08 the different local behaviors,
00:26:18.15 and allows evolution to take place
00:26:22.02 without complicating evolution somewhere else.
00:26:25.08 So, you can vary the...
00:26:29.09 your digits from...
00:26:31.14 you know, as I say, from the...
00:26:34.13 almost the digits of a whale to the digits or a human hand
00:26:37.16 to the digits of a bird
00:26:40.26 because that region of the embryo
00:26:43.02 is told to deal differently
00:26:45.09 with exactly the same information
00:26:47.07 as another region of the embryo.
00:26:50.11 Well, the last
00:26:52.20 -- and I think, maybe, arguably the most important --
00:26:54.23 is the most biochemical
00:26:57.05 and maybe a little more abstract,
00:26:59.15 and this is about the ease
00:27:02.05 with which you can change the regulation
00:27:04.04 of what's being expressed in the...
00:27:08.18 in a given region, or how much is being expressed,
00:27:13.05 or when it's being expressed.
00:27:15.26 And it's a property of molecular systems
00:27:20.20 called weak linkage.
00:27:23.22 So, what is weak linkage?
00:27:25.18 It's a form of regulation that's easily established and easily changed.
00:27:28.26 So, if you want to think about weak linkage,
00:27:30.22 you could think about the electric plug
00:27:33.10 and the electric outlet.
00:27:35.16 You know, this... it's an ancient system.
00:27:39.08 The plug and the outlet haven't changed in,
00:27:41.21 you know, 100... in 100 years.
00:27:43.28 But what's plugged into that outlet,
00:27:46.24 and what's generating the power,
00:27:49.20 has changed radically.
00:27:52.01 So, it's a very conserved form
00:27:53.28 which has this property,
00:27:56.10 this easy change.
00:27:58.15 The regulatory signal,
00:28:01.16 which is... in some system...
00:28:04.05 provides little information and provides...
00:28:07.25 it works, really, by permissive rather than instructive interactions.
00:28:11.29 That is, the system already has several states,
00:28:15.06 and it just lets one state happen or another state happen
00:28:19.19 rather than telling it exactly what to do.
00:28:21.23 And this property of many biochemical a
00:28:24.12 nd molecular systems
00:28:26.21 allows easy evolution of new combinations.
00:28:31.01 So, what you have here, in general,
00:28:34.11 more abstractly,
00:28:36.14 is a system that has itself
00:28:39.03 an on and an off state.
00:28:41.04 And the signal is just something that inhibits it
00:28:45.08 either in the on or the off state.
00:28:47.17 And the evolution of inhibitors
00:28:49.25 is very easy in biological systems.
00:28:51.23 All you need to have for an inhibitor
00:28:53.22 is just something that binds to one of those states preferentially,
00:28:55.26 and you stabilize that state in a different state.
00:28:58.29 And this works in metabolism,
00:29:00.22 it works in transcription,
00:29:02.16 it works in all sorts of behaviors in the cytoskeleton.
00:29:05.20 And so, I think... I won't go into in detail,
00:29:08.16 but that just tells you it's very easy to plug new things
00:29:12.05 in and out of outlets,
00:29:15.00 because that system already exists.
00:29:16.29 It's... it allows for change,
00:29:19.11 and we can easily select
00:29:23.06 what we put in and what we...
00:29:25.01 what we... on both ends. Alright.
00:29:27.13 So, of these... all these properties...
00:29:30.08 you know, the regulation, the compartmentation, the weak linkage...
00:29:37.08 all of these things provide evolvability,
00:29:41.02 make it easier for novelty to arise.
00:29:45.06 And I would argue that living systems are built
00:29:49.17 such that minimal genetic change
00:29:51.11 can cause large changes in the organism,
00:29:53.15 so that says it's not what Darwin thought,
00:29:57.27 that it's not that the change is so infinitesimally small.
00:30:00.29 It can be quite large,
00:30:02.18 and therefore can happen more rapidly.
00:30:04.12 But these are...
00:30:07.05 generally are produced without lethality,
00:30:09.18 because these systems are very tolerant,
00:30:11.18 and they tend toward utility,
00:30:13.17 because in some similar way
00:30:15.15 they worked well in the past
00:30:18.02 and incorporated some of those features.
00:30:20.20 So, here we are in the modern molecular age.
00:30:24.25 And we would like to understand novelty
00:30:30.14 down to its molecular roots,
00:30:33.07 at the level in which genetics acts,
00:30:35.14 which is down at the single amino acid level,
00:30:37.21 at the single nucleotide level,
00:30:39.22 and drugs act, which is down at the...
00:30:42.08 small regions of proteins.
00:30:48.03 And we have a long way to go.
00:30:52.14 So, I'm gonna try in this last part of the talk
00:30:54.15 to suggest a strategy
00:30:57.24 by which we may be able think about how to go further.
00:31:00.18 So, I want to stress -- unlike things I said before,
00:31:02.21 which I think are well established by example --
00:31:04.24 this is more trying to predict
00:31:08.02 how we might approach such a problem.
00:31:10.11 Okay.
00:31:12.14 So, we're very proud of all the molecular data
00:31:14.22 that we've accumulated over the years,
00:31:16.18 and we can represent the connections
00:31:19.10 between different genes in an organism as a graph --
00:31:21.29 that is, what activates what, what inactivates what --
00:31:25.02 but it's a very complicated and presently uninterpretable graph.
00:31:29.05 And it's... and it's only a sort of binary graph.
00:31:32.01 I mean, it doesn't tell you the strengths of the interactions.
00:31:35.07 It doesn't tell you the timing of the interactions.
00:31:37.07 There's a lot of things it doesn't tell you.
00:31:39.08 And this typical genetic interaction map
00:31:41.18 is sometimes called a hairball
00:31:44.09 because it's so informative.
00:31:46.21 So, how are we gonna get beyond...
00:31:49.13 we had Darwin are telling us that genes were important --
00:31:53.00 he didn't know about genes, but at least that genetics was important.
00:31:56.08 We have... we have what I've talked about, evolvability,
00:32:01.00 all these conserved processes
00:32:03.05 which can be plugged in differently,
00:32:05.23 can be localized, and can be...
00:32:08.09 and can generate different states quite readily.
00:32:11.29 They can explore different systems
00:32:13.22 and compensate and suppress lethality.
00:32:15.14 We have all that.
00:32:17.14 How do we get from the age...
00:32:19.12 and then we have, of course, selection.
00:32:21.00 How do we get from the genes to the phenotype?
00:32:25.16 And this is probably...
00:32:27.21 a lot of people think it's the holy grail for all of biology,
00:32:30.17 not just for evolution --
00:32:32.10 that is, understanding the gene to phenotype map.
00:32:35.20 And so... and that would be useful,
00:32:40.23 not just, as I said, for evolution,
00:32:43.00 but it would tell us how the environment molds our bodies,
00:32:47.12 our minds, our behavior, and our health,
00:32:50.26 and it would define the changing conditions
00:32:53.14 in which evolvability can do its magic.
00:32:56.06 So, we might extend their lives,
00:32:58.11 improve our well-being, remove stress,
00:33:01.00 or take the clues from what we've learned
00:33:04.11 and make changes in our environment,
00:33:06.09 or even, now, in our genome.
00:33:08.19 Now, how to go forward from the complexity
00:33:11.13 of information we're building up,
00:33:13.05 and the goal,
00:33:15.03 which is some sort of understanding,
00:33:16.21 which means some real reduction in the dimensionality
00:33:19.15 of this information
00:33:21.29 to something we can perceive?
00:33:24.12 Is there a possible way to go forward?
00:33:26.20 So, let's deal with two amazing, complex processes,
00:33:31.05 which I would argue have something in common,
00:33:35.15 at least formally.
00:33:37.09 One is how random processes
00:33:39.20 that evolved the human being
00:33:43.17 in just three billion years of evolution.
00:33:45.27 Many people find that remarkable.
00:33:48.05 Some people find it so remarkable
00:33:50.16 they think it's supernatural.
00:33:53.08 I think of similar remarkability
00:33:57.25 is how a baby or a young child
00:34:00.17 in three years can learn a language
00:34:04.11 completely foreign to him,
00:34:05.28 not encoded in his brain in any way.
00:34:08.02 And of course, that brain is not predisposed
00:34:09.23 to learn any given language.
00:34:11.14 It can learn any language.
00:34:14.01 And that is a problem
00:34:16.24 where the study has been quite different in form
00:34:20.22 than the study in biology, of evolution.
00:34:24.00 Can we learn something about what approach to take
00:34:28.11 by comparing the similarities of these two problems?
00:34:32.17 Now, I would argue there's a striking parallel
00:34:34.24 between evolution and cognition,
00:34:36.19 at least on the formal level
00:34:38.15 -- not necessarily on the actual molecular level.
00:34:41.22 In evolution, we've talked about novelty, evolvability,
00:34:46.21 exploratory processes, compartmentalization,
00:34:49.15 weak linkage, regulation;
00:34:55.19 and on the cognitive level,
00:34:58.03 we can see that the parallel to novelty is learning.
00:35:03.05 The formal parallel of evolvability
00:35:06.05 is learnability.
00:35:09.06 It's how... you know, how is something written in a book
00:35:13.16 so that you can understand it?
00:35:15.13 It's similar to how,
00:35:17.20 you know, the biological system is set up
00:35:20.16 so that it can easily change in a useful way.
00:35:23.12 Exploratory processes are called reinforcement in cognition.
00:35:27.17 Compartmentalization has been called modularity in cognition,
00:35:31.06 weak leakage, plasticity,
00:35:33.11 and regulation, rewiring our brains.
00:35:36.19 Now, the people studying cognition
00:35:41.24 have applied a lot of computer-based approaches
00:35:47.02 and have made some progress on that.
00:35:49.07 A new development
00:35:51.21 is the understanding of cognition
00:35:54.04 through reverse engineering and deep learning.
00:35:59.21 Tremendous amounts of detail are in the picture of...
00:36:02.25 any given picture,
00:36:04.29 but it has to be reduced to something which can be then
00:36:08.28 compared with some known detail,
00:36:10.21 which may not be exactly the same,
00:36:12.22 to recognize something as a cat, for example,
00:36:14.27 in a picture.
00:36:17.03 The same is true in a self-driving car, which is...
00:36:21.19 has some principles of operating,
00:36:23.17 but the detail is always changing.
00:36:25.22 Or the ability of a machine...
00:36:33.03 a computer to beat the top Go player in the world,
00:36:35.27 where there's a nearly infinite number of possible Go games
00:36:38.20 that could not be memorized and reconstructed.
00:36:42.02 Some operating model has to be generated,
00:36:45.09 and that's what we're looking for
00:36:47.26 in evolution.
00:36:49.14 So, the question is,
00:36:52.09 can reverse engineering and deep learning
00:36:55.17 help us understand how evolvability operates
00:36:58.25 throughout biology?
00:37:00.05 I don't know the answer,
00:37:01.20 but I think it's worth trying.
00:37:03.17 Can they tell us something
00:37:05.10 about the normal function and growth, health, and disease?
00:37:07.26 Can they point to a suite of important molecular changes
00:37:11.01 in evolution?
00:37:13.03 I think this is something we have to try.
00:37:16.02 But we have to recognize that this is a difficult process
00:37:18.23 and we need certain requirements
00:37:21.05 in order to apply these kinds of methods.
00:37:25.05 So, the answer to the question is,
00:37:27.07 yes, but...
00:37:29.03 and there are challenging steps ahead.
00:37:32.04 We need to be able to measure multiple phenotypes
00:37:35.00 in some way, such as...
00:37:38.17 and measure their growth, their differentiation,
00:37:41.20 and their behavior.
00:37:43.23 We probably should be learning at small...
00:37:46.05 focusing on small graded perturbations,
00:37:49.25 for example, using drugs or by studying steps
00:37:53.12 in embryonic development,
00:37:55.13 or by looking at very similar species.
00:37:58.18 We should probably be studying these things
00:38:00.25 in a natural setting,
00:38:02.17 for example, not... tumor cells are so dysregulated
00:38:05.29 that they're probably not good for studying how normal cells function
00:38:10.07 in these complex circuits.
00:38:12.16 And we need quantitative information,
00:38:15.07 high quality molecular measurements.
00:38:17.19 And it is simply not true
00:38:21.08 that a large amount of big, noisy data is going to work.
00:38:23.23 So... and from that,
00:38:26.14 we need to be able to derive simulatable computational models.
00:38:30.09 So, I'd like to sort of summarize
00:38:36.27 just where we are and what we may have learned.
00:38:40.20 Evolvability, I think, as a concept,
00:38:43.03 has changed the way we think of evolution,
00:38:45.08 the cell, and the organism.
00:38:47.06 It's dealt with the variational component
00:38:50.00 in phenotype.
00:38:51.26 And it's the variation which is selected upon.
00:38:55.12 The selection doesn't generate the variation.
00:38:58.05 So, we need to know how variation is generated.
00:39:01.07 Though evolvability helps explain evolutionary change,
00:39:04.16 the picture is still blurred
00:39:07.06 by the lack of even a crude map of genes
00:39:09.19 to the organism's phenotype.
00:39:12.22 And the formal similarity of novelty to learning
00:39:18.07 I think offers some hope.
00:39:19.29 And we may soon be able to use new computational approaches
00:39:23.06 to better understand the nature of evolvability,
00:39:26.05 the nature of evolution itself,
00:39:28.05 robustness, and, most likely, disease.
00:39:31.11 Thank you.

This Talk

Speaker: Marc Kirschner

Audience:

Educators of H. School / Intro Undergrad
Student
Educators of Adv. Undergrad / Grad
Researcher
Educators

Recorded: July 2018

Talk Overview

Marc Kirschner begins his talk by defining evolvability as the capacity of an organism to generate phenotypic novelty, particularly novelty that is useful and heritable. He goes on to explain that evolvability may be driven by just a few mechanisms. For example, changing the time or place in which a gene is expressed may be sufficient to generate a novel phenotype without requiring a change in the gene product. Until we have a better understanding of how genes map to phenotypes, however, it remains difficult to understand the nature of evolvability. Kirschner suggests that perhaps advances in computational learning will help us to decipher how evolvability works through biology.

Speaker Bio

Marc Kirschner

In 1993, Dr. Marc Kirschner joined Harvard University where he became the founding chair of the Department of Cell Biology. In 2003, he moved to Harvard Medical School to found the Department of Systems Biology. Research in Kirschner’s lab focuses on problems that require the coordination of biological events in time and space. His lab… Continue Reading