Evolution in Action?

From researchers at Harvard, here is stunning time-lapse photography of something that (as far as I can see) might or might not be evolution. A bacteria colony spreads out across a giant Petri dish, hits an area with a high concentration of antibiotics, appears to be stopped in its tracks, but then slowly breaches the boundary in a limited area and is soon spreading as before until it reaches an area with an even higher concentration of antibiotics, and the story repeats:

According to the narrator, the barrier is being breached by a small number of antibiotic-resistant mutants, which then reproduce like crazy, at least until they come into competition with other mutants, etc. In other words, evolution.

Now I’m sure this is an extremely naive question, but I hope someone can offer an answer that will leave me less naive: How do we know this is evolution as opposed to learned behavior?

Suppose a given bacterium has several strategies for fighting off the effects of toxins, only one (or a few) of which are succeessful against this particular antibiotic. Most bacteria choose ineffective strategies and die; a few others, by random chance, choose strategies that work. Eventually, other bacteria “observe” that Strategies A,B,C,D,E,F and G keep failing but Strategy H appears to work — and therefore stop using Strategies A,B,C,D,E,F and G and switch over to H, allowing the colony to continue expanding, etc. How, exactly, do these observations take place? I (at least in my current state of naivete) can easily imagine that the bacteria long ago evolved the ability to make such observations, say by detecting chemical evidence that Strategy H has been deployed in certain places where the colony appears to be growing, and that other strategies have been deployed in areas where the colony appears to be dying. I can even imagine that they’ve evolved a mechanism to “deliberately” leave chemical evidence of their choices to make these observations easier for their relatives.

If that’s the story, the bacteria that take over their world at the end of the video can be genetically identically to the ones we see at the beginning. In other words, no evolution takes place over the course of the video.

But the researchers seem quite sure that this is a video of evolution, presumably because they know things I don’t know. Can someone share that knowledge?


34 Responses to “Evolution in Action?”

  1. 1 1 Val

    The shortest answer would be that everybody else died and somebody magically didn’t, and that somebody then became the new default.

    (I wouldn’t be at all surprised if you could find a paper associated with the video that mentions having sequenced various sets of the bacteria; it is surprisingly cheap to do nowadays.)

  2. 2 2 Steve Landsburg

    Val: I agree that gene sequencing could settle the issue. But the claim in the video seems to be that the visual evidence is all we need. That can be true only if there’s some a priori reason to reject the alternative theory.

  3. 3 3 SeanV

    Well, the alternative scenario that you describe could in fact be the case. If the deposit of a chemical trace could be detected by other relatives and not by unrelated bugs then it could work in the way you describe. Also, if it benefits a bug to have lots of other even unrelated bugs around then it could perhaps also work. Otherwise there would be no obvious advantage to doing so. It’s interesting because a bug and another bug would have to simultaneously have to evolve both the “leave a trace” adaptation and the “copy any bug that leaves this particular trace” contemporaneously (or pretty close).
    The counter argument could be that we simply don’t observe such signals – although they may not be looking for them. But also, that even if it does happen – bugs depositing chemical traces – then it’s still an evolutionary adaptation that occurred via random chance, etc, at some point in the past – and we’ve move the event back so to speak.
    Social insects do stuff like leave chemical traces for the benefit of others but that has a simple evolutionary explanation based on the peculiar genetic relatedness between individuals.(don’t forget ant colonies are genocidal!)
    Human culture (and other animals) is of course an example of knowledge transmission but is not believed to have occurred much in bugs.
    Interestingly of course parasites have an ability to make themselves resistant to immune system attack by changing the molecular structure of their “coat” (sex is believed to have evolved to counter this). So this is a kind of “learning”.
    Also, the immune system “learns” and can adapt in ways that are not pre-programmed – as does the neural system. So, in those modules the ability to adapt is what is inherited rather than a fixed adaptation itself.

  4. 4 4 David Pinto

    My understanding of this from biochemistry class at Harvard is that resistant bacteria contain a plasmid (small circle of DNA) and endows resistance to antibiotics. There is a cost to having this extra bit of DNA, it takes longer (more energy) for these bacteria to reproduce. In the absence of antibiotics, non-resistant bacteria dominate.

    So what you are seeing in natural selection. Existing structures in the creature give it an advantage in certain situations. This is what Darwin observed in the birds on the Galapagos Islands. They were all the same creature, but different features became dominant depending on pressures of a particular island.

    It’s not evolution in terms of watching mutations occur.

  5. 5 5 Steve Fritzinger

    They don’t say it explicitly in the video, but the dots and tree structures in the last 20 seconds look like they were sequencing the colonies genomes and could see exactly where a resistant daughter population split from the parent.

    A Harvard Gazette article on the experiment talks about tracking mutants as they migrate across the mega-plate. The Science paper is still pay-walled, so I didn’t check it. But talking about mutants supports them secuencing the genomes.


  6. 6 6 Alex

    The language here is not dispositive, but it’s very suggestive:


    “But the MEGA-plate isn’t just a fancy visual aide. It’s also a valuable research tool. Baym and his colleagues can collect microbes from different places on the plate and sequence their DNA. They can then reconstruct the gradual accumulation of mutations that allowed some bacteria to make it all the way from the safe periphery to the deadly centre. They can work out which mutations matter.”

  7. 7 7 HAO

    Researchers can analysis the bacterial DNA or RNA. Mutants should have different genes compare to the original one. Hence evolution.

  8. 8 8 Amit Hajra

    You can’t tell the difference from the video. You would have to sequence the DNA of the resistant bacteria. Presumably, the initial group of resistant bacteria will all have the same DNA variation (mutation) compared to the original group. The second resistant group will also have this mutation but will have an additional mutation that will make them resistant to the higher dosage. So if evolution is at work, there should be at least three genetically distinct groups of bacteria at the end of the video (there could be more because multiple mutations may lead to the same resistance) You are correct that if the bacteria are all genetically identical, then it is not evolution at work.

  9. 9 9 Dave

    Did they check the DNA of the “before” and “after” cultures?

  10. 10 10 Jens Fiederer

    It seems to me the problem with the “learning” interpretation is that the bacteria are never shown this video. Their knowledge must be based on strictly local observation. If there were “scout” bacteria that roamed around and reported back what strategies worked best, learning would be more plausible.

  11. 11 11 Biopolitical

    The original paper gives a detailed account of the mutations appearing in each of the successful and unsuccessful bacterial lineages that grow, or fail to grow, forward on the plate. These mutations alter the function of enzymes involved the metabolic pathways disrupted by the antibiotics used in the study. There is a lot of previous work on those enzymes and their mutations and their role in antibiotic resistance.

    The video alone does not let us discriminate between evolution and learning. Even the evidence in the paper is not enough to discard learning. But then we have Occam’s razor.

    Here is the paper: http://science.sciencemag.org/content/353/6304/1147.full

  12. 12 12 MWC

    A few things to note:

    The ‘strategies’ available to them are fairly limited, especially in prokaryotic cells. I’m not sure what the time scale is, but I’m guessing it’s over several hours/days. Other biological signaling mechanisms act on the order of seconds/minutes. The mechanism of action of the antibiotics is well known, and bacteria aren’t known to ‘respond’ to them. Because they only move by replicating, there’s no way to signal to everyone else that they’ve ‘chosen’ the right strategy. It’s way more consistent with random acquired mutations that provide improved fitness to the progeny.

  13. 13 13 Scott F

    So if this would be ‘learned’ behavior, there would I’d think have to be a some kind of information exchanged. That information would have to be exchanged along some medium (no pun intended).
    Some options:
    Some bacteria have receptors that are sensitive to light. As far as I’m aware, they don’t use that ‘visual’ information to do anything either than have more or less exposure to light, but perhaps it’s more complex than that.
    This may be a mundane version of what you’re talking about, but horizontal gene transfer is entirely a thing in bacteria. Some bacteria accomplish this through sex pili, others through picking up pieces of “naked” DNA called plasmids. But this is a kind of learning.
    Hydra can also ‘learn’ tricks taught to other hydra in a similar fashion.

  14. 14 14 Ken B

    The lesson would have to be learnt by the children, wouldn’t it? And grand children. Doesn’t that require the learning is durably chemically encoded? The only plausible candidates for durably encoding such information would be DNA and RNA. This is an empirical finding; DNA and RNA can encode information persistently in a way proteins cannot because of the evolved replicative machinery. They are the only game in town.

    Under your scenario there would need to be a chemical, Stevene, which is present *and maintained* in higher quantities in generation N than in generation 1. How does generation N maintain a higher concentration of Stevene? It must be *expressing* Stevene at a higher rate. If this is not a result of a change in DNA or RNA, but in some other protein which stimulates or previously repressed Stevene, Grothendieckene, don’t we have a recursion?

  15. 15 15 Biopolitical

    Ken B,

    First, all learning, in humans or in any other organism, involves chemical processes that are not “encoded” in DNA or RNA. Second, there exist so-called epigenetic processes that transmit information (of the sort of “express Stevene at a high rate”) to progeny, and that don’t involve learning and don’t require changes in nucleotide sequences.

  16. 16 16 Ken B

    Biopolitical, if you find bacteria writing books you might have a point. As it is you have missed mine.
    You are also wrong on epigenetics, where no effects beyond a couple generations have ever been evidenced. See Coyne’s articles on this.

  17. 17 17 nobody.really

    Under your scenario there would need to be a chemical, Stevene, which is present *and maintained* in higher quantities in generation N than in generation 1. How does generation N maintain a higher concentration of Stevene? It must be *expressing* Stevene at a higher rate. If this is not a result of a change in DNA or RNA, but in some other protein which stimulates or previously repressed Stevene, Grothendieckene, don’t we have a recursion?

    Not necessarily. Are we needlessly precluding the possibility of a miracle? I’d heard that Harvard would be exposing blameless bacteria to antibiotics. And I’m pro-life. So I prayed to Mother Teresa, and behold!

    (But I didn’t pray for the bacteria in the control group; they’re apostates.)

  18. 18 18 Joe

    I propose Occam’s razor as an answer to your question.

  19. 19 19 Steve Landsburg

    Ken B:

    The lesson would have to be learnt by the children, wouldn’t it? And grand children. Doesn’t that require the learning is durably chemically encoded?

    Not at all. Bacteria have strategies A,B,C,D,E,F,G. They can release chemicals Stevene-A, -B, -C, -D, -E, -F, and -G. An individual who successfully fights off the antibiotic with strategy X celebrates by emitting a little Stevene-X. An individual who encounters a little Stevene-X employs strategy X, succeeds, and releases a little more Stevene-X, which is found by the next individual to come along, etc.

  20. 20 20 Scott F

    @ Steve 19 I think this is in part the strategy used for quorum sensing and biofilm creation in bugs like S. aureus. When enough bacteria are around, each producing some kind of here-I-am molecule. When they reach some threshold of there-are-manyness they’ll start to secrete the proteins necessary to form a biofilm. This is a sort of strategic response to having ‘numbers’ on your side.

    But it seems counter-intuitive to me that this would happen in the case of the the E. coli above, because It would be in the best interest of the E. coli that figures out how to live on antibiotic concentration [x] to want to minimize the number of competitors it has on this new fresh space that it has adapted to (I’m assuming that there is some advantage to moving into the new space, or why do it at all). Thus, sending out Stevene-X (the molecule that transmits the information on how to survive at concentration[x]) would be maladaptive. It might even make sense for the E. coli to send out misinformation.
    It seems to me that one could get a sense of which information sharing strategy the the E. coli are using by watching where successive ‘plumes’ of E. coli start after the first adapter appears. The information sharing hypothesis would be supported by nearer-than-random plumes showing up after the initial plume. Misinformation could be typified by more-distant-than-random plumes occurring after the initial plume. Or the classic random genetic adaptation hypothesis would be supported, if the successive plumes were completely unrelated spatially.
    One would probably need to take into account that once the bacteria start to interact with (and potentially metabolize) the antibiotic, it effectively lowers the concentration in the local area, decreasing the extent to which another E. coli must adapt to start dividing int the next zone.

  21. 21 21 Ken B

    @Steve 19
    Hmm. Let’s look at the mechanics a bit more closely.
    Exta-cellular Stevene-X stimulates expression of Stevene-X. This is your idea I think.
    Why wasn’t Stevene-X production running out of control earlier, due to its presence in the cell already? Because it is modified extra-cellularly, but still serves as a stimulant perhaps? Stevene to Kenene stimulating Stevene. Conceivable I suppose if a bit baroque.
    (Have you been reading Wynne-Edwards?)

    I can imagine an empirical test.

    Take Stevene-X-taught bacteria from generation N, and isolate it for several generations, N of them say, in a new medium so that the Stevene-X has dissipated.
    Under your theory a population of these distant descendants should have to relearn, but under mine (durable chemically encoded learing in the genome) they shouldn’t. Stevene-X will be their native tongue. Agree?

  22. 22 22 roystgnr

    Ken B, your test depends on how fast unused antibiotic resistance is selected *against*, doesn’t it? If you make N larger than some N_chem then the chemical-trace-using bacteria will have mostly run out of Stevene-X and will again be vulnerable to the antibiotic, but if you make N larger than some N_gene then the Gene-X bacteria will evolve again to no longer have that gene and will also still be vulnerable to the antibiotic.

    I suspect that N_gene is much much larger than N_chem, because diffusion is fast and genetic drift (or even a weak negative selection against the cost of a useless strategy X) is slow, but I have no idea how to guess at the size of either.

  23. 23 23 Steve Landsburg

    Biopolitical (11): I don’t see any obvious application of Ockham’s Razor here. Is natural selection always simpler than learning? Does Ockham’s Razor suggest that the reason we have the Internet is that all the people who didn’t have the Internet died off?

    On the other hand, the question does seem to be settled (rather spectacularly) by the paper you linked to.

  24. 24 24 Ken B

    @22 roystgnr
    Yes, if there is very rapid selection against the X gene then a later generation might have to “start over”. It seems unlikely. How did the X gene manage to persist up until this experiment was done? That’s SAteve’s theory right: X gene and others are there now. It would have to be some sort of frequency selection, like sex ratios I think.

    THIS IS WHERE OCCAM’S RAZOR COMES IN. My theory postulates a well known mechanism, genetic change. Steve’s is a parlay.

    Empirical results though are never logically water-tight; refer to comment 17, which is logically impeccable.

  25. 25 25 maznak

    If there were some reasonably small number of strategies to choose from, or “learn”, you would observe that the barrier is breached simultaneously at many places, not only one – as there are millions (or billions) of bacteria presumably trying different strategies at the same time.But the original barrier seems to be breached on only one place, which is more consistent with a very unlikely event (mutation) happening to one randomly selected individual. The further breaches look like “once the mutant is resistant, it is almost just as resistant to 10x higher concentration”. This surprised me most, that the next barriers looked like almost no barriers at all.

  26. 26 26 Harold

    This is not evolution in the sense of becoming more complex, which is where the creationists have an issue. It is a demonstration of natural selection and inheritance of characteristics. Essential components of evolution, but not, I suggest, the whole bundle.

    The reason we reject learning is, as others have said, because we would need to invent a whole new mechanism for which there is no evidence. There is oodles of evidence that it works the way we think it does. Of course the evidence in the video is not enough in itself, but it does not stand alone. It fits in with all the physical evidence we have acquired.

    We could say that the film of the feather falling as fast as the metal object on the moon is not proof that gravity acts the same on all objects and it is air resistance that slows the feather on Earth. Maybe in vacuum feathers want to fall faster, so race ahead, but air induces an indolence in feathers that results in them falling slowly on Earth. We would discount this as silly because we have so much other evidence that does not require stating that the gravity explanation is the better one.

    Other observations. Each side has three mutants that make it into band 2. On the right, the first central mutant is the only one with progeny that make it all the way to the middle. On the left, three different mutants have representatives at the center. The right side is also faster, presumably because the central mutation is more effective. If allowed to go to completion I presume only progeny of this one would survive from all 6 strains that made it past the first band.

    If we extended the experiment with more central bands, but used a different antibiotic in the new central bands, we may find that the left side did better because it has three possibilities to work with whereas the right side has only one. Maintaining greater diversity could well provide an advantage to new environmental stresses. Thus in easy times lots of variation can be preserved that may come in useful in the future. Some say that human evolution has stopped or slowed because we can keep all genotypes alive. Apart from this being demonstrably wrong, if so it would simply be phase of increasing diversity which will provide greater survivability when the balloon goes up.

  27. 27 27 Dave H.

    The initial question assumes that bacteria have different strategies. As I understand it, they do not. Antibiotics prevent the bacterial cells from multiplying, most frequently, by stopping the mechanism responsible for bacteria building their own cell walls. Bacteria have several “methods” for building cell walls, but not several “choices.” (Each individual will typically only have one. The wall building blocks consist of sugar molecules linked to a lipid carrier that anchors them to the cell membrane. The wall is built using a process common to all bacteria.)

    At the individual level, there is no such thing as a “learned” behavior. There are only two behaviors: Adapt or die.

    Although it is true that antibiotic resistance can either be inherent or acquired, both count as “evolution”. The resistance trait is being passed down genetically in either case, regardless of whether it was a novel mutation or a pre-existing characteristic.

  28. 28 28 Ken B

    “The initial question assumes that bacteria have different strategies. As I understand it, they do not. ”

    You know, I doubt this very much. Different alleles in a locus can correspond to different proteins or different regulation, and these represent different strategies. Humans have different strategies for eye color and molar shape. There is no a priori reason to think that a population of bacteria might not have a distribution of strategies for dealing with an anti-biotic.

  29. 29 29 Harold

    There are four basic strategies for resistance:
    drug destruction, such as lactamase enzymes against penicillin
    alteration of target binding site
    alteration of metabolic pathway
    reduced drug accumulation, by preventing access or speeding removal.

    Each of these basic strategies will have many variations.

  30. 30 30 Ken B

    @Harold 29
    I have no complaint about your comment, but in the sense Steve meant “strategy” refers to any different approach: any of your variations (or indeed, combination of them).

  31. 31 31 Harold

    Yes, I agree. Steve’s proposal is that the bacteria have all these strategies available to them from the beginning and must choose between them. There are no new strategies. Further, it assumes that the strategy is not simply selected but communicated and adopted by different bacteria.

    Even if all bacteria already had every strategy available to them there is no need to require learning as long as the progeny expressed the same as the parent. Given the speed of reproduction and the speed of communication it is probable that learning would be unnecessary in this situation as the progeny very rapidly multiply. So we can dispense with the learning argument and we are left with evolution, or merely selection.

    Selection certainly seems possible from the video alone. The peppered moth is often cited as an example of evolution in action. The moth was predominantly light coloured to blend in with its background. With the soot from the industrial revolution the background became dark, the few dark individuals survived, and most of the population became dark. With the clean air recently, the opposite occurred and we now have mostly light colored moths. This demonstrates natural selection which is essential for evolution, but it does not demonstrate evolution itself. Similarly with the bacteria. Are we watching selection, or evolution?

    The scientists seem sure that we are observing new strategies resulting from mutations. It is not possible to know this from the video alone.

    On the learning side, bacteria have plasmids, small strings of DNA that they swap among themselves. Bacteria in some sense do learn; by swapping plasmids, successful strategies are passed among a population and even between species.

  32. 32 32 Ken B

    Steve is questioning that there is a genetic change, and your learning via plasmid swapping falls on “my” side of the debate. If there is a persistent chemical memory encoding the learning it is memorialized in Dna or rna say I. “Not so fast” says Steve.

  33. 33 33 Harold

    I think there is little doubt that there is genetic change, but this cannot be seen from the video alone.

  34. 34 34 Advo

    I think the obvious answer is that bacteria don’t have anything TO LEARN WITH.

Comments are currently closed.