Sanjuro

The punch line:

When forced to stand on their own two pili, obligate cheater Myxococcus will sometimes evolve into a non-cheater that outcompetes all of its progenitors and contemporaries.

The details:

Six replicates of an experiment were carried out that combined an obligate cheater of M. Xanthus, called OC, with a non-cheating cousin called GJV2 in a 1% OC/GJV2 mix.  Each replicate was cycled through alternating growth and starvation (called Development for some reason) stages for a total of 6 cycles (i.e. G1-G6 and D1-D6).  Eventually only one of the replicates was followed.  In this replicate, the population had crashed to near unmeasurable levels by D4.  OC was essentially acting as a DI particle, and overwhelming the GJV2 clones until the population crashed.  However, the surviving clones in D4 had apparently evolved the ability to sporulate properly and not only outcompeted GJV2, but were also immune to OC’s efforts to cheat.  This new super Myxo was called Phoenix (PX).

 
The authors found that PX was different from OC by only one nucleotide, in the middle position of a 7 run homopolymer (i.e. a 7 base run of a single type of nucleotide, in this case cytosine).  This run was 128 bases upstream of an acetyltransferase gene, of which there are approximately 30 in the Myxo genome.  It is still not clear exactly what effect the mutation is having.  Acetyltransferase levels are not consistently higher in PX than in other progenitors, though the pattern of expression is certainly different.  It is not clear whether the mutation is only affecting acetyltransferase expression, if it affects a range of other proteins, or if the change in acetyltransferase expression could have knock-on regulatory effects. 

 
An additional experiment was carried out in which the non-cheating parent of OC was transformed to include the PX mutation.  The result was that the OC parent displayed the PX phenotype.  It would really be interesting to see is if the original GJV1 and GJV2 would show the PX phenotype if transformed in the same manner.  There are an additional 14 single nucleotide changes that took place between GJV1 and PX and if they transformed GJV1, they would be able to see if the mutation that created PX was dependent on the other mutations that occurred between GJV1 and PX.  It might also be useful to know what changes might have taken place between GJV1 and GJV2, just to see if any of the same mutations arose separately.

After discussing the Cooper paper in class, and thinking about it some more, I am still not convinced that the data presented in the second part of the paper supports the FM model.  During discussion I learned that I didn’t fully understand the predicted effects of recombination.  My interpretation was that recombination would serve to drive many mutations to fixation at roughly the same time, or at least linearly as they occur.  It seems that prediction is actually that the time to fixation will be accelerated by recombination, not just sped up.  This means that any beneficial mutation that occurs will have a lasting and increasing advantage (in terms of frequency, not fitness) over any other mutation.

The portion of the results I still struggle with are best summarized in Figure 3.  The rec- graphs are as expected, the initial competitive advantage conferred by SpoT is challenged by subsequent beneficial mutations.  However, the results from the rec+ lines do not seem to fit either my multiple fixation mistake, or the accelerated FM model prediction.

My multiple fixation mistake model would predict that any initial competitive advantage (in relation to contemporary clones) conferred by a beneficial mutations at low generation numbers would be less at high generation numbers because of the rapid spread of other beneficial mutations.  In other words, the single (or few) mutations(s) providing the advantage early on, would be overwhelmed by the greater number of mutations shared by all clones at a later stage.  If this was the case, we would expect there to be either positive or negative relative fitness numbers in the beginning (probably positive since the clones are unlikely acquire two strongly positive mutations during the early stages).  At the end of the experiment, we would expect to see relative fitnesses close to 1 across the board as more mutations spread through the clones.  This pattern appears in the comparisons between timepoints 3 and 4 but not 1 and 2 or 5 and 6.

In the acceleration model, we would expect to see a small fitness advantage in the beginning that grew to be larger at the end because of the effect of acceleration.  We see this pattern at timepoints 1 and 2 but not in 3 and 4 or 5 and 6.  However, we could imagine exceptions to this pattern.  For example, suppose the small advantage provided in early generations from spoT fixed quickly because of recombination.  However, consider the possibility that another, more advantageous, mutation arose subsequent to the first comparison (we have presumably seen such mutations in the rec- lines).  This new mutation would spread quickly because of recombination and it is possible that a second, later, comparison between spoT+ and spoT- clones would show that the spoT+ clone had a comparative disadvantage to the new spoT- clone (that included the second mutation).  This example shows that the pattern of relative fitness that might arise from repeated tests could be very complicated and varied, and that many comparisons would be required to provide information about the underlying evolutionary pattern.  In the paper we are given only three examples and even those examples do not point to any one explanation consistently.

I may obviously be missing something else but I’m still not convinced that the data provided in the paper, particularly the rec+ data, provide support for the mechanism by which recombination helps reduce clonal interference.

I don’t know why I’ve had so much trouble with the Cooper paper but I have.  The first part seems clear enough, E. coli that can recombine outcompete those that can’t.  From figure 2 it also seems clear that beneficial alleles fix faster in the recombining lines than the non-recombining lines.

The second part of the experiment is what I don’t understand.  He does several comparisons between different combinations of time points, spoT +/-, and rec +/-.  The short story seems to be that high generation spoT+ rec- lines do not have a fitness advantage over high generation spoT- rec- lines.  This is in contrast to high generation spoT+ rec+ lines which retain their advantage over high generation spoT- rec+ lines (I am still struggling to figure out what comparisons he is making, the combination of diagrams and text is not specific enough for me to figure it out).  What confuses me is why this is a confirmation of the FM model.

I would think that the FM model would predict more similar fitness between different rec+ lines and more variation in fitness between rec- lines.  In rec- lines, the assumption is that each line develops its own beneficial mutation and that these different mutations prevent the fixation of any one beneficial mutant.  If this is true, then the relative fitness of any high generation rec- line compared to another high generation rec- line would be dependent on only one or two beneficial mutations (or upon the accumulation of deleterious mutations).  For rec+ lines, because they can recombine, we would expect successful lines to have multiple beneficial mutations.  Differences in fitness between rec+ lines would therefore be dependent on multiple mutations.  If this is true, then we would expect relative fitness between different high generation rec+ lines to be less variable than relative fitness between different high generation rec- lines.   Cooper seems to be making the point that the consistently higher fitness of high generation spoT+ rec+ lines supports the FM model but I’m not sure how it does that.

I also don’t understand how the experiment supports the hypothesis that the accumulation of competing beneficial mutations, as opposed to the loss of deleterious mutations, are responsible for the fitness gains.  At several points Cooper asserts that the presence or absence of deleterious mutations can not explain the data, but I don’t understand his explanation.

I’m looking forward to the discussion.

I had this tape deck once where the rewind button was broken and I had to flip the tape over and hit fast forward if I wanted to get to the next song (for you young whippersnappers, tape decks were used to play cassette tapes, where the music was recorded on thin magnetized plastic strips. In order to get to the next song you had to fast forward, stop, listen, fast forward some more, stop listen, rinse, repeat. I remember the day the feature came out that allowed you to fast forward to the next song, which was great except for those Mike Oldfield albums where there were no real breaks in between songs. But I digress.). I think the rewind button on evolution may be broken too.

As I was reading the paper I was liking the experimental design, it seemed simple and straightforward and teased apart the various forces that could shape evolution. But when I sat down to start writing this, I started to wonder about what the experiment was actually telling us. The authors have arrived at some sample values for the relative contributions of adaptation, chance, and history, but these values are highly dependent on the conditions under which the experiments are run. For instance, what if conditions were used under which a smaller cell size was an adaptive advantage. You would get very different values for the three forces of evolution. The contribution levels derived from any one experiment can not really be generalized (and to be fair to the authors, I don’t think they assert the the contribution levels are constant).

But if relative contributions are not constant, is their technique useful? I don’t think you could rewind much of tape of life using this technique, it is too complicated and far too deep. However, perhaps it is possible to carry out this study on recently diverged populations. Let’s say there is a recently created subpopulation of a species and you want to determine the role of the different evolutionary forces on the formation of this subpopulation. Perhaps you could do an experiment like this to make such a determination. This might be useful for ecologists trying to understand the forces shaping the diversity within an ecosystem. But as far as helping understand the relative role of adaptation, history, and chance in shaping millions of years of evolution, I’m a bit skeptical.

I suppose the best blog I had was about the SAR11 paper, although now that I have thought about it more, I’m not really sure I hit the nail on the head. I’m still not sure why they used synteny as a measure, was it just to find the hypervariable regions? And how syntenic was it really? Is there a known distribution of synteny that allows us to say this is relatively syntenic and others aren’t?

But I think the real value of the paper for me was that it drove home Abe’s point in his Borg post. I have been a bit confused about how sequence similarity could provide information about relatedness. The lesson of the SAR11 paper is that sequence similarity alone can not be used to say two things are related. Each one of the SAR11 genomes studied is more related to the other SAR11 genomes than it is to anything else, which indicates that, in spite of the sequence divergence, they are evolutionarily closely related. Trying to provide an arbitrary threshold for sequence similarity would miss the real story.

So my SAR11 post combined with this one should be considered my best.

Ok, this site has finally pissed me off to blog about it. I’m happily composing a comment and when I hit the post button I get an error message that says I have not provided my name. There is no option to go back to the comment and when I hit the back button on the browser, I get a blank template again. Very frustrating. So here is a short list of things that frustrate me about the site. Sorry Vaughn, the lost comment is making me sound more annoyed than I should, I’ll try to be constructive.

1. As mentioned above, it would be nice for the error page to preserve your post in some way, especially since you can’t save comments like you can regular posts.

2. It would be nice to have an RSS or ATOM feed for the entire class, rather than each individual member. It would streamline my RSS reader and make it easier to see when something new had been posted.

3. Many of the Themes are broken.  I had to go through quite a few to find one that was working and that had a login link on it.  Some sort of standards and testing would be great.   Maybe impossible given the open nature of the site.

4. It would be helpful to have a little bit more structure to the collection of blogs.  None of the individual blogs link back to the main class blog unless you specifically put a link in.  This just may not be part of the overall vision of how the site should work, but having indviduals associated with a class so that the class blog functioned as more of a unit would make navigation easier.

5.   My personal preference would be to have the blogs actually be threads on a bulletin board.  This would provide the structure mentioned above, feeds, easy flagging of new comments, and would facilitate threaded discussions.  This is just my own preference I guess.

Well, gosh, don’t I just feel better after that rant.

This SAR 11 paper has me running in circles a bit. But there are a few threads of thought that keep coming up that I’ll go over.

Synteny. Its role in this study is a little bit strange. Unless I’m missing something, the one real benefit of looking at syntenic regions is that they have been able to identify HVRs, which in turn allows them to bolster the whole ‘core’ genome hypothesis. I’m still a bit unsure about the level the synteny they have actually measured. Stephen brought up in class that a syntenic call could be made with just two genes. In addition, the description for Figure 2 indicates that there could be 5 gene deletions allowed and the fragments could still be called syntenic (I’m assuming this means that genes A and G are together in the metagenome but are separated by genes B-F HTCC1062). Also, the e-value used as the cutoff is very low, necessarily so because the amino acid sequences are so diverged. All of this makes me wonder whether the level of synteny detected is highly dependent on the parameters you use (e.g. different e-value cutoffs, different criteria for what constitutes synteny). That being said, I’m not sure how much it matters here because the authors do not seem to provide any explanation for why synteny has been maintained.

The other thing I wonder about is this high degree of synteny with the low sequence similarity. I may be stuck on this whole idea of species but if we rule out our prior knowledge that these are all members of the same species, would we ever call two organisms that had such a low degree of sequence similarity the same species? The data is fairly convincing given the fact that the genes from all of the data are reciprocal best hits to each other. So, since the metagenomic SAR 11, MTCC1062, and MTCC 1002 are indeed part of the same species, however we want to define that, am I right in thinking that in order to have such high levels of functionally similar genes, there would have to be a very high level of homologous genetic exchange taking place between strains of Pelagibacter ubique? If there was not a high level of exchange taking place, then the selective forces that operate on the HVRs, which are presumably the local adaptations, would lead to local bottlenecks in populations that would eventually lead to divergence between local strains of P ubique. Only a high recombination rate would allow there to be relative uniformity (by function, if not by sequence similarity) among the core genes while maintaining high levels of divergence at the HRV regions. This should be testable by looking at P. ubique that are not so geographically removed from the Oregon strains. We would presumably be able to find some genes that are close the Oregon strains in terms of sequence (thinking back to the mosaic of Whitaker).

Related to that, am I right in remembering that synteny usually degrades faster than sequence similarity does? If this is the case, is it because of non-homologous recombination? I’m assuming that homologous recombination would likely lead to sequence variation without syntenic degradation. If this is the case, then P. ubique would also have to be very resistant to this kind of non-homologous recombination, more so than other bacteria.

And finally, is it possible that P. ubique are so widespread precisely because they have these (unique?) characteristics of high levels of homologous recombination combined with high resistance to non-homologous combinations? And following from this, is it possible that the variation we see in P. ubique is just as much due to their recombining characteristics as it is to their “infinite” population size? Or, in other words, is their high degree of diversity, which we would predict in a very large population, in part due to unique characteristics that allow them to be diverse without diverging. I really need to read that HGT review.

Ok, I think I’ve gone about as far out on the limb as I can possibly go. I’m done.

These are a bit scattered, a combination of notes and comments.

p375 last para – Restating what they said, in a large population drift has less effect because of the time taken for fixation? I think it’s right that the chance of a new allele occurring and fixing in a small population is the same as in a large one. Therefore they are talking about the time taken for fixation. That is, if HGT is rare and drift is slow, then the mutation rate would outpace the rate of fixation and variation would be extreme (in the absence of periodic selection).

p377 first para – Detection of divergence in the Stable Ecotype Model is highly dependent on the assumption that there is very little HGT between ecotypes. Is the evidence that Whitaker et al found for high rates of recombination specific to Sulfolobus? Are there certain types of bacteria where the ability to exchange DNA is positively selected for? Are Sulfolobus the exception or is HGT more common in more types of bacteria than this paper would suggest?

Discomfort with genetic exchange only model:
What if a bacteria does not exchange genetic material very often (i.e. with any other bacteria, including its close “siblings”) but is virtually identical to another bacteria and is closely related. Yet this would be considered “less related” than bacteria that can occasionally exchange genes but that are very different from each other, both by sequence and by phenotype.

Discomfort with the ecotype model:
Is it possible that ecosystems on opposite different parts of the globe could have similar ecotypes and have very different bacteria that have evolved from different lineages to inhabit those similar ecotypes? If the answer is that they are not the same ecotypes because they are in different locations, what about oceanic environments where you may have the same bacteria (as measured by sequence similarity or readiness of genetic exchange) in seemingly similar ecologies but far removed by geography? Abe and I have also talked about the problem of convergent evolution. Many Eukaryotic traits have evolved independently in different lineages (eyes for example). If two bacteria evolve to inhabit the same ecotype, would this be a problem for the theory?

Questions:
–What are the problems with defining the concept of species for Eukaryotes? Could there be the same problem for bacteria.
–Interbreeding as a criteria is necessary but not sufficient in Eukaryotes. This may not be true in prokaryotes.
–Can Periodic Selection also be thought of as instantaneous drift? The selection event is certainly “choosing” certain alleles (or combinations of alleles) but there are lots of other alleles of lots of other genes that are fixed or eliminated purely by chance, because of their association with the selected allele.

I guess I need to read the papers that we discussed on Tuesday.  There were a few questions I had about the experiment and what they found.

1.  How can we be sure that it was a mutation and not environmental stimulus that changed the morphology of the bacteria?  This could be easy to test by changing the environment and seeing if they change to a different morph but is it possible that once a certain shape has been taken in response to environmental stimulus, that it can’t be changed?  This could clearly happen in eukaryotes in the development stages but I don’t know enough about bacteria to know if this is possible.  In other words, one a certain set of genes has been activated, is it possible that part of the process of activation is to permanently turn off other genes.

2.  A more technical question, when they take the bacteria out of the vial and put them on the plate, is there any danger that they are skewing the ratios because the different forms grow differently when plated?  Is it possible that there is some ascertainment bias here?

3.  If the changes in morphology are indeed due to mutation, what is the size of the mutational target?  If the target is small, is it just that there are so many individuals that you are bound to see the mutation?  But if so, isn’t it possible that the mutation would happen so late that you would not find significant numbers of one form or the other?  30 generations isn’t much, although 1 billion individuals is.

The paper probably has answers to all of these things, I just haven’t looked yet.