Parallel adaptation in fish: same genes used over and over

April 11, 2012 • 6:03 am

When a species encounters a new environment on multiple occasions, but the same environment, how often does it re-use the same genes when adapting to the similar but novel habitats?  This is an unanswered question in evolutionary biology because such parallel adaptations can evolve by two genetic routes. First, the ancestral species might adapt simply by using genes that are already present (albeit at low frequencies) in the main body of the species—the so-called “standing genetic variation.” Every species, including ours, has a pool of genetic variants kicking around in its gene pool, many of them “neutral” or perhaps slightly deleterious but maintained by recurrent mutation. If the environment changes, these “standing” variants can be used as fuel for adaptation, so one might predict that the same variants might be used again and again when adapting to the same circumstance (so-called “parallel evolution”).

The other way to adapt is to use “new” mutations that happen to arise after one encounters a new habitat. In this case one might predict that multiple cases of adaptation to the same habitat might use different genetic variants, because many mutations are possible, they arise randomly, and selection will pick up only those that happen to arise.  (One exception here: if there are only one or a few possible mutations in the genome for adapting to the environment, then selection might have to wait for those to occur, and in the end you’ll also see the same mutations used again and again. This might explain why widely different species of insects adapt to organophosphate insecticides using mutations at the very same gene—an enzyme that detoxifies the insecticide—indeed, at the very same amino acid codon in that gene.) You can distinguish the two routes of adaptation by DNA sequencing: if the same standing variants are used repeatedly, they should show fundamental similarities in their DNA sequences, since they all derive from a common ancestral mutation.

A new paper in Nature by Felicity C. Jones et al. (there are a gazillion authors, though the last is the amiable David Kingsley, an evo-devo guy at Stanford who is the motive force behind most of this genetic work on sticklebacks), favors the former hypothesis: the re-use of standing genetic variation in multiple independent cases of adaptation.  The species is the threespine stickleback (Gasterosteus aculeatus).  The ancestral form of this fish was marine, but it has invaded freshwater—streams and lakes—multiple times, and each time it did so the fish evolved in similar directions.

So there are now two types, the marine (which does return to freshwater to breed) and the freshwater.  All the marine fish look pretty similar: they are larger, and deeper bodied, can survive better in salt water, and, most notably, are covered with bony armor and have pelvic spines. The freshwater forms are smaller and thinner, not tolerant of salt water, almost completely lack the bony armor of the marine forms, and many (but not all) such populations also lack the pelvis and the pelvic spine (hindfin).  The difference in armor plating and spiny-ness almost certainly reflects the relative lack of predators in the freshwater environment as opposed to the open ocean: it’s adaptive to not have to waste metabolic energy making protective structures you don’t need, and that’s supported by the fact that, when raised in a common environment, the marine form grows more slowly than the freshwater form.

Here’s a diagram showing the differences between the forms (top is freshwater, bottom marine; remember that these “forms” are classified not as different species, but as different subspecies of G. aculeatus). Note the divergence in armor plating, body size and shape, and the positions of the fins, the spines, and other “morphological landmarks” (diagram prepared by Dr. Felicity Jones of Stanford, first author of the paper, used with permission).

This more striking photo of the skeleton shows the big difference in bony armor and spines between the forms (remember that not all freshwater populations lack spines and pelvises, though all lack armor):

You can see a very nice lecture by David Kingsley on the evolution of sticklebacks here.  I recommend watching this one-hour video, produced by the Howard Hughes Medical Institute.

The Jones et al. paper used a clever technique designed to answer one question: since the freshwater form has evolved repeatedly throughout the world from the marine form (probably via some ancestors remaining in freshwater after breeding instead of returning to the sea), and has evolved parallel morphologies each time, how often is the genetic basis of this adaptive change similar? That is, how often has this parallel evolution used the same genetic variants?  The answer is “pretty often.”

The group used two clever ways to figure this out, which involved the tedious sequencing of 21 entire stickleback genomes: a reference genome and 10 pairs of fish from nearby localities, with one member of each pair being marine and one pair being freshwater. (The freshwater populations they studied lacked armor, but not spines or pelvises.) In essence, the method did a “family tree” of genes, and looked for those genes that all grouped together by habitat, i.e. were genetically similar among all freshwater forms, with the variants of those genes grouping in all saltwater forms. (Looking at most genes gives a “traditional” grouping, which groups fish by geographic region [i.e., Atlantic vs. Pacific] as opposed to habitat: that’s exactly what you’d expect if the evolutionary history of the forms was one of repeated colonization of freshwater habitats from saltwater.)

They found four nice results:

  • The two methods each identified over 240 regions of the genome that responded in parallel when marine fish adapted to freshwater. That’s about 0.5% of the total genome.  Using only those genetic regions found in common with the two methods, the authors found 147 candidate regions for parallel adaptation (0.2% of the genome).  One of the regions that evolved in all freshwater fish contains a form of the autosmal Eda (hypohidrotic ectodermal dysplasia) allele, which is involved in formation of scales, plates, teeth, and other surface structures in fish and mammals.  This locus was implicated in previous work by the Kingsley lab, and the two forms that of the gene that make armor versus no armor differ in both coding and non-coding positions of the gene, so we’re not sure whether the difference in armor is due to a structural or regulatory change in the DNA.  The lesson, though, is that there has been substantial and large-scale use of the same genetic regions in parallel adaptation.  This is probably, as Kingsley has posited before,  because these variants (especially in Eda) are present in low frequency in all worldwide marine populations, and were simply present fortuitiously to serve as fuel for adaptation when climate change induced some marine populations to become permanent inhabitants of freshwater.
  • The genetic regions implicated in parallel evolution contain (or are close to) contain genes involved in “cellular response to signals, behavioural interaction between organisms, amine and fatty acid metabolism, cell–cell junctions and WNT signalling,” as well as kidney function, as expected for fish that have to live in freshwater vs. saltwater.  And of course the Eda gene was involved, as was known previously.
  • Many of the genes occurred close together on chromosomes, i.e., they were physically linked.  Through clever sequencing techniques, Jones et al. determined that this physical linkage (on chromosome 1, 11, and 21) was caused by the presence of chromosomal inversions: sections of rearranged chromosomes that contain the selected genes.  The marine forms have one form of chromosomal architecture; the freshwater forms have another.  The association of newly evolving genes with such chromosome rearrangements is adaptive, because the inverted regions keep adaptive gene complexes together (a cross-over, or recombination, between different inversions leads to the production of chromosomes with duplicated or deficient regions, which cause death in their carriers. Recombinants having mixtures “wrong” genes, then, like kidneys for freshwater and armor for saltwater, can’t survive). The association of inversions with adaptive genetic differences is a sign that the marine and freshwater forms hybridized during or after their differentiation, for the associations will not be favored unless the forms have a chance to mate with each other during their divergence.
  • Finally, the authors divided the genetic regions involved in parallel divergence into three classes: “coding regions” (those genome regions that involve a difference in protein sequence), “regulatory regions” (those regions where there is parallel adaptation but which contain no structural genes, so that the adaptation probably results, as in Eda, from differences in gene regulation that affect the expression but not the sequence of the protein product), and third, “probably regulatory” regions (regions that contain coding and noncoding sequences but where the authors couldn’t pinpoint a structural protein difference that distinguished all marine from all freshwater forms).  Jones et al. looked at 64 regions identified by both of their methods as having the strongest signs of parallel evolution, and here’s the breakdown of these 64 genetic regions involved in parallel adaptation:

As you can see, only about 1 in 6 of the evolved regions involved a change in protein sequence, while the rest involved likely or potential changes in gene regulation.  In this case, then, parallel evolution seems to have occurred largely through changes in when and how genes are expressed, not in the sequences of the genes themselves.  This goes along with evo-devotee Sean Carroll’s suggestion that evolutionary changes in animal form will largely involve changes in gene regulation rather than sequence.

Three caveats here, though.  First, Carroll’s hypothesis applied to animal form and not physiology (though he never explained why), and yet the changes we see in this study involve both form and physiology.

Second, the changes that Jones et al. studied were those involved in parallel adaptation: similar changes in multiple environments.  The chart above doesn’t tell us anything about the spectrum of genetic change involved in unique adaptations: those that were specific to one or some but not all instances of freshwater invasion. We know there are such genes, for Jones et al. describe some.

Third, other studies don’t show such a high concentration of regulatory regions vs. structural regions in the evolving genome. As Jones et al. note:

Mutations causing structural changes in proteins are the most abundant variants recovered in laboratory Escherichia coli and yeast evolution experiments. They make up 90% of 40 published examples of adaptive changes between closely related taxa, and 63–77% of the known molecular basis of phenotypic traits in domesticated or wild species. The larger fraction of regulatory changes implicated during repeated stickleback evolution may reflect our use of whole-genome rather than candidate gene approaches, stronger selection against loss-of-function and pleiotropic protein-coding changes in natural populations than in laboratory or domesticated organisms, or an increasing prevalence of regulatory changes at interspecific compared to intraspecific levels, including emerging species such as marine and freshwater sticklebacks.

To my mind, the data are still out on the Carroll hypothesis.  Five years ago, Hopi Hoekstra and I published a paper in Evolution (reference below) questioning whether the “rush to judgment” about the importance of regulatory genes in evolution was really warranted by the scanty amount of supporting data. Our point was not to push for the important of structural-gene changes in evolution (though they surely must be important, for many new genes, like globins or immune-system genes or olfactory genes, arise by the duplication of previously-existing genes followed by evolutionary divergence, and those are surely structural changes), but to call for more data to resolve the issue.  We were misunderstood, I think, as saying that evolution largely proceeds by structural gene changes, and people like Sean Carroll got mad at us. But at the time there simply weren’t enough data to resolve the issue.  I still don’t think there are, but I’m perfectly happy to accept a predominance of regulatory-gene changes in adaptive evolution if that’s what the data wind up showing.

The data of Jones et al. do show this for parallel evolution, and we need more work along those lines to resolve the issue. (That work, though, should involve both parallel and unique cases of adaptation.) But what is clear from their paper is that standing genetic variation can be used over and over again if a species has to adapt independently to similar habitats.  And that tells us something about how evolution worked, at least in this species.

This is a superb paper: congrats to Felicity Jones and to David Kingsley and his group. I love seeing long-standing evolutionary questions answered by a combination of classical morphological data and high-tech genetic work.


F. C. Jones et al. 2012.  The genomic basis of adaptive evolution in threespine sticklebacks. Nature 484:55-61

Hoekstra, H. E., and J. A. Coyne. 2007. The locus of evolution: Evo devo and the genetics of adaptation. Evolution 61:995-1016.

35 thoughts on “Parallel adaptation in fish: same genes used over and over

  1. Jerry, as you know, the duplicate globin genes you cite — “many new genes, like globin … genes, arise by the duplication of previously-existing genes followed by evolutionary divergence, and those are structural changes” — were necessarily accompanied by regulatory changes as well, such that different gene duplicates are expressed at distinct phases of individual development. I’m not sure that Sean would recognise them as counterexamples.

    1. Yes, I believe Hopi and I pointed that out in our paper. Certainly fetal globins, for instance, are expressed at different times from beta globin. But it’s equally certain that changes in gene STRUCTURE are important here too, as the proteins do different things. The point is that we know there is no one overweeningly important form of evolution, and the task is to figure out how often structural vs. protein evolution are important. In this csse both are, in other cases only structural variation is important (e.g. insecticide resistance), while in others regulatory variation is important (Eda). The problem is that the “regulatory” people, I think, don’t talk much about gene duplication in general.

      1. On re-reading my post I wish I had started with “Avid reader, big fan, thanks for this post and all others, but I have a minor quibble…” Best wishes and thanks,

      2. I’m doing my PhD on globin regulation, gene duplications, regulatory vs structural evolution etc… Early days at the moment, but keep an eye out for some ground-breaking papers on these topics 😛

  2. I haven’t read the paper in full yet but I can’t see whether they have determined whether copy number variation is important.
    If you are looking at the genomes of any two individuals of the same species it is the CNVs rather than point mutations in either coding or non coding regions, that makes up the bulk of the sequence differences. In my own field of cancer genetics it appears that whole genome sequencing gives a very incomplete picture of the changes that are occurring – you really need a combination of genomic sequence, CNVs and transcriptional changes (as well as the epigenome and proteome) to figure out what is driving the phenotype you can see.
    In the case of the sticklebacks it would make sense that regulatory changes rather than coding region changes are involved. This allows much greater opportunity for switching between states when the environment changes – rather than a coding alteration where the ‘fix’ might involve a very rare specific change. Regulation, on the other hand can be changed by multiple elements (transcription factor sites or levels, microRNAs, ubiquitin ligases, mRNA stability, etc) any of which could be a target of subsequent mutation.

    1. It’s a matter of perspective. CNVs and other structural variation such as segmental duplication are really just mutations at the genomic or chromosome level. Gross chromosomal rearrangement are hallmarks of certain cancers and other human diseases and are now just beginning to be appreciated for creating DNA variation in primate evolution.

      Say a single duplication event separates humans from chimps. Another duplication within that first duplication (found in a small percentage of humans) impairs brain development.

  3. Interesting work for sure, but this genome-level stuff is maddeningly vague to a physiologist. Candidate regions are close to genes involved in kidney function? Explain!
    In my own parochial view I see this work as preliminary– as the authors (all of ’em) say in the article:

    The many marine–freshwater divergent loci and gene expression changes identified in the current study will substantially accelerate ongoing searches for the genetic and molecular basis of fitness-related morphological, physiological and behavioural differences between marine and freshwater fish.

    These more specific, functional studies will be most interesting.

  4. note that these “forms” are classified not as different species, but as different subspecies

    Do these forms interbreed, and are there intermediate forms? They seem so radically different that I’m surprised they are not considered separate species.

    1. There are hybrid populations; the paper gives an example of one. And there are hybrid zones between the forms. That, I suppose, is why they aren’t designated as separate species: they interbreed and produce fertile offspring in nature.

  5. I read it too although my genes and life experiences have joined together in a massive conspiracy against me understanding much of it

  6. One marine species is giving rise to several/many independent freshwater forms ± simultaneously in different river systems. Incipient species all other the place, but not a dichotomy in sight.

  7. Just checking in to reassure Jerry that people are reading the science articles.

    A thought I just had: does wordpress have a widget or something you can put on a page for a reader to indicate that he or she read it without actually commenting? If so that might be a better tool for judging interest in an article than the number of comments (especially since the number of comments is very frequently a few people commenting a bunch of times).

  8. This paper is of the same sort as another recent paper, on dogs.

    Genome Analysis of the Domestic Dog (Korean Jindo) by Massively Parallel Sequencing.

    The Jindo is the korean version of the Dingo (though since pedigreed). The breed was mapped to a boxer reference genome. 70,000 insertion-deletions (3 of which were within genes involved in cranial-facial morphology) and 8,000 genome structural variants were identified between the two. Other variation included olfactory receptor genes and the mitochondrial genome.

  9. Very interesting post.

    Its about time someone writes a book which puts this marvelous phenomenon into a context consistent with 21st century science.

  10. Interesting. Quite apart from the genetics, I’ve learnt about armour-plated marine sticklebacks!

    “evo-devotee” – cute.


  11. Regards 4/5 of evolutionary changes (as the sticklebacks changed environment) pertaining to regulation of protein expression rather than to the what protein is coded for; is this ratio generally the same for all vertebrate evolution, as well as everything else (e.g. bacteria)?

    Also, I take it most coding mutations simply modify an existing allele; in comparison does someone know how frequently completely-new genes are added? (And of those, what ratio arise from a structural mutation that duplicates a gene into two chromosomal loci which then slowly diverge, rather than a small mutation transforming a noncoding region into a coding region?)

    1. I don’t know about the rest, but doesn’t the article answer your first question plenty? I.e. it is still uncertain?

  12. Nice paper – just read this one for a reading group! I think it would be really interesting to know what’s going on as far as gene flow and hybridization among marine/freshwater population pairs. i.e. what’s maintaining all those freshwater-adapted alleles in the ocean?

    1. As someone from the “stickleback community”I can tell you there is hopefully a lot more work going on to answer these questions. The problem is the ecology/life-history of the marine populations is actually not that well known. It isn’t known whether marine populations return to their natal locations to breed year after year (like Atlantic salmon) and it isn’t clear where they go at sea either.

      However, the leading hypothesis on freshwater alleles in the marine population is that continual gene flow between marine and freshwater populations maintains the alleles at low frequencies. The Eda allele for complete plates for example is dominant and modifier loci also help maintain a marine phenotype. Since these recessive alleles are masked from selection in heterozygous individuals, they are able to spread geographically. Gene flow is pretty substantial between freshwater and marine populations in some cases and the marine populations show comparatively low genetic structure. My bet is on these sink-source dynamics… the question is how to test it?

      Also Jerry… big fan of the blog, the book and Speciation. Glad to see you writing about my organism of choice also!

      1. If the regulatory exploitation of “standing genetic variation” turns out to be the preponderant adaptive mechanism in nature (which may be a premature judgment, admittedly), then surely this discovery undermines the role of ‘random variation and natural selection’ as an adaptive mechanism? It seems to me that it must, since it conjures up images of organisms that are pre-equipped to cope with environmental changes (or at least a certain amount of these) in fairly well-defined and predictable ways. The ancestral marine sticklebacks seemed predestined for further diversification, or for certain physiological/anatomical changes, so their adapting to freshwater environments seems more inevitable than random…one naturally starts to wonder to what extent the marine ancestors could really have failed to deliver these changes in freshwater environments?

        1. Short R’s, remembering I’m an armchair commenter:

          – “Undermines”.

          I’m not sure what you say here. The article argues for that there is no undermining, but that it is an open question how much each mechanism contribute. They have evidence for both here, see the diagram.

          – “Pre-equipped, pre-destined”.

          Well, those who hadn’t the necessary variation may not be around to tell us, so this looks to me like selection bias on your part.

          Also, I thought it was generally accepted that there isn’t, and can’t be, a destiny in the overall process? It is environmentally dependent. (Unless you mean the destiny of our biosphere to die off as our sun moves towards the read giant phase.)

          – “more inevitable”.

          I think it would be very difficult for some of them to not be able to do so, considering that extant fishes are claimed to derive from freshwater fish. So the transitions ocean-sea-ocean has already been observed. Presumably the possibility of the first ocean-sea transition made it darn near impossible not to be able to transit back, bar serious lock in effects.

          And indeed I believe these transitions occur regularly, and some fishes do it several times during their lifetime.

          The question wasn’t if this was possible or difficult as I understand it. The question was that it is easy enough for _two_ mechanisms to be responsible, so it was a question of ratio (and maybe the amount of dependence between them).

  13. Finally got around to read this. It was well worth the effort, the first two paragraphs taught me more evo (or perhaps evo-devo) than the rest of the year.

    Speaking of the relentless need for more, since fishes seems to surface on evolution so much, while kittehs are, well, kittehs, I’m looking forward to my first reading of catfish post here.

      1. Thanks for the comments,Torbjorn. I will read the article again and perhaps reply at a later stage.

Leave a Reply