Highlight text to annotate itX
David Kingsley: So thanks very much, and it's a huge pleasure
to be here at NHGRI when so much has happened and is happening in the area of genomics.
I'm actually a product of the eighth grade museum experience as you heard about from
Kirk, Kirk Johnson. We all know animals differ in all sorts of very interesting traits, and
I'm very fortunate to get to work on old problems related to what are the actual molecular changes
that produce those interesting differences that you see between animals.
So there's -- that's a very hard problem. Although there's lots of organisms that have
been sequenced, we still can't look at a sequence and tell you what the animal looks like. And
that problem is related to a whole series of other issues, including how many genetic
changes actually underlie the interesting differences. If you see some interesting trait,
is it lots and lots of little tiny changes that produced it, or smaller number of changes
in major genes? What are the most important genes? What types of mutations occur in those
genes? And if you give evolution some problem to solve, are there lots of different ways
of doing it, or does evolution tend to use particular mechanisms over and over again?
So about 15 years ago, we got interested in trying to study those old problems by trying
to bring a genetic approach to the study of evolutionary change in vertebrates. We went
looking for some organism that had undergone very dramatic, and very recent and repeated
evolutionary experiments in nature. We chose a small fish, the three-spined stickleback,
which lives in the ocean, but they're like salmon. They migrate into the freshwater streams,
and lakes, and coastal areas to breed every spring. That migratory lifestyle set off a
huge evolutionary radiation when the glaciers melting, created all sorts of new environments
in North America, Europe, and Asia. They got colonized by marine sticklebacks, and since
had 10,000 years, or approximately 10,000 generations, to adapt to the new food sources,
and water conditions, and predators, and salinities, and water colors, et cetera, in these new
environments around the world.
So if you go around freshwater lakes and streams, tens of thousands of new environments created
since glacial melting, you see really dramatic differences between these newly-established
populations: huge changes in jaws, and teethes, and spines, and boney armor, and pelvic apparatus,
and color, and physiology, and behavior, all of which have evolved in just the last 10,000
years. So although many of these forms look as different as you would see in different
generative animals, it's only 10,000 years of evolution, and the isolating mechanisms
between the forms tend to be behavioral and mechanical in compatibilities between populations.
Those can be overcome using artificial fertilization in the laboratory. So if you squeeze out eggs
and *** from very different looking sticklebacks, and do fertilization in Petri dishes, you
can actually raise fertile F1 hybrids. That makes it possible to then raise large families,
and to treat evolutionary change in natural populations as a genetics problem, can we
actually map the key chromosome regions that control the differences, and eventually get
down to the genes and the mutations that are responsible for these major changes.
I have to say, when we first got interested in this, sticklebacks had great biology, had
been studied for decades, thousands of papers, several full-length books written about the
traits and ecological significance of different traits, but they weren't an organism that
had attracted interest from molecular geneticists, so there was essentially no sequences, markers,
maps, transgenics; none of the things that you need for a real model organism. We spent
a lot of time with support from NHGRI over the last 10 years building four three-spined
stickleback exactly the same sorts of tools we found useful in mouse genetics; we're trying
to track traits all the way from morphology down to chromosome genes and mutations. And
so what I'll do today is illustrate how we're trying to use many of those tools to study
some of the cool, interesting differences that have evolved repeatedly in these natural
So let's start with limb formulation, a great example; limb modifications, of course, one
of the major changes that you see in organisms adapted to different lifestyles. One of the
biggest changes in limb pattern in different vertebrates is the complete loss or reduction
of limbs that's evolved in many animals, including loss of both forelimbs and hindlimbs that's
evolved in snakes; loss of only the hind legs, which has evolved in marine mammals including
both whales and manatees; and that selective loss of the hindlimb has also evolved in both
fossil and many living stickleback populations.
Here's a reconstruction of the skeletal structures of a marine stickleback, heavily armored.
I've colored the pelvic hindfin of the fish here in red. Sticklebacks are like most fish
and land animals. You have two set of paired appendages: pectoral fins or arm homologs,
pelvic fins or leg homologs, and the pelvic fin of the stickleback consists of this bony
spine that articulates with an underlying pelvis. The fish can raise that bony spine
as a defense against soft-mouthed predatory fish that try to eat sticklebacks in the ocean.
So although ocean sticklebacks always have a robust pelvic apparatus, many of the ocean
fish have colonized environments where there are no other predatory fish. Some of those
lakes are shallow, full of insects; insects actually pray on sticklebacks by grabbing
onto the dorsal fin pelvic spines, and reeling them in, and eating them from the sides, and
in that sort of environment, there is actually a selective advantage to losing the pelvis,
which has occurred in a couple of dozen of populations around the world.
So that actually gives us the chance to try to look at what's the genetic architecture
of this sort of dramatic limb modification that's evolved repeatedly in natural population.
So to study that, we cross marine fish with a robust pelvic apparatus to a lake fish that's
completely lost the pelvis, raise a couple of thousand offspring, isolate DNA from the
F2 animals, type them with genome-wise set of linkage markers, and compare it to pelvic
size. And when we do that, what we find is pelvic production isn't the simple Mendelian
trait, but it's not a miserably complicated infinitesimally tons-of-regions trait either.
There's a single chromosome region that controls about two-thirds of the variation in pelvic
size in the cross as well as a series of unlinked modifier genes.
Well, one of the nice things about a map is that it allows you to quickly evaluate any
candidates that are known. There's a lot known about limb development, and when we mapped
the stickleback homologs of a whole series of limb genes, we found that this particular
gene called Pitx1 maps directly to the major chromosome region that controls the size of
the pelvis in the cross. That's actually a very interesting candidate gene because it's
a homeodomain transcription factor that plays an important role in regulation the expression
of lots of other genes. It has this name, Pitx1, based on a group that isolated it,
studying pituitary development. But the same year, it was isolated by another group who
named it the backfoot gene. It was named the backfoot gene because it has this very striking
expression pattern where it's found in hindlimbs, but not the forelimbs, of the whole series
of animals, all the way from fish, to birds, to mice, and humans. The gene's been knocked
out by two different groups: the knock-out mice have small hindlimbs. They actually die
at birth with pituitary defects and jaw abnormalities, cleft palate. So in many ways, that doesn't
look like a very promising basis for evolving new traits in natural populations, which brings
us to the view of the neo-Darwinian synthesis about the likelihood that major genes are
going to underlie adaptive traits in nature.
The quotable Ernst Mayr here talking about the fact that most mutations found in laboratory
organisms that have dramatic effects on phenotypes are actually deleterious in reduced fitness.
He said that to believe these sort of drastic mutations would produce viable adaptive phenotypes
is equivalent to believing in miracles. And you can see the problem; I mean, this is a
good example of a mouse that has obvious viability problems, although also a component of the
phenotype is also something that we see in nature.
So what's actually happened in the stickleback? Well, when we sequence Pitx1 in marine and
freshwater fish, what we find is that coding region of the gene is identical in both types.
However, when we look at expression patterns, there's an interesting difference: larva from
marine populations show Pitx1 expression in the lips and inside the head in the pituitary,
and then this little spot of expression along the side, that corresponds to the site where
that pelvic hindfin would normally develop. Freshwater fish still show expression in the
head and inside in the pituitary. They completely lost this little site of expression along
the side of the body. That isn't just because they've lost the pelvis. You can cross the
marine and the lake fish; the F1 hybrid actually forms the pelvis. But is the pelvis, the only
allele that's expressed comes from the marine fish, not the freshwater fish, so that suggests
there's been a cyst-acting change in the regulatory apparatus of the Pitx1. The freshwater sticklebacks
have preserved the coding region, they've conserved expression for jaws in pituitary,
but they've lost the ability to turn this gene on in one particular spot in the body.
We think that can happen by the lost of a modular regulatory element that drives expression
at that location.
We've been looking to see whether that sort of model can be worked out in detail. Is it
true? High-resolution mapping and lab cross in the natural populations has shown that
the genotype in a 20 kb region upstream of the Pitx1 gene can predict, in a natural population,
the presence or absence of a pelvis. We've built a whole series of transgenic constructs
using marine sequences from that region, hooking them up to GFP reporters, and Frank Chan injected
them into fertilized stickleback eggs, looking for little magic pieces of DNA that would
drive expression specifically at that site of the body. And they're there. I hope you
can see this, it's a fairly specific expression pattern, but there's a small fragment of DNA
in that key candidate region that drives GFP reporter expression specifically at the site
that the pelvis will normally form in the fish. If this is really the right gene and
that's really the right sequence, you also ought to be able to do a more ambitious experiment,
and that's to try to reverse the evolutionary change.
So for this, we hooked the marine control region not up to a GFP reporter but to a Pitx1
cDNA, and we inject that into the eggs of an evolved pelvic-less population that hasn't
made an external pelvic spine for thousands of years, and we were thrilled to see that
the introduction of the marine information will actually put the pelvis back on the stickleback.
Here's the original vestigial apparatus that's seen in the evolved freshwater population,
and here's one of these transgenic fish, where the introduction of the marine information
stimulates the formation of a nice serrated spine that articulates with a restored pelvic
apparatus on the ventral side of the fish.
Okay, so that, then, with our hands on the right gene and piece of DNA, we can finally
look at what has actually happened in different populations around the world. One of the nice
things about the stickleback is the phenotype evolves over and over again. Series of complementation
and mapping, and then C2 experiments implicate the same major chromosome region again and
again when pelvic reduction evolves at different locations. And when Frank Chan looked at DNA
sequences from a whole series of different pelvic-reduced populations around the world,
what he found was a whole series of deletion of a few hundred to a few thousand base pairs
that completely eliminate this pelvic enhancer region.
Okay, so I went through that in some detail to give you some idea of the actual genetic
and molecular basis of a major skeletal alteration that's evolved in these natural populations
for pelvic reduction. If you actually look at the results from the crosses, you see that
relatively few chromosome regions can have big effects. The chromosome region with the
biggest effect corresponds to the location of a key developmental control gene that's
required for the formation of multiple different tissues. Despite its essential role, regulatory
mutations in that gene provide a way of producing a very large phenotypic alteration at a specific
site in the body, and that's such a good mechanism for evolving this phenotype that when pelvic
reduction evolves over and over again in different populations around the world, exactly the
same mechanism is being used over and over again with a repeated series of deletions
that eliminate that regulatory element.
Well, we've been interested in using similar methods to study a whole range of traits.
One of the reasons we thought it was justified to build a genetic tool kit for these fish
was because of the large range of different phenotypes that have evolved in different
populations. To give you another quick example from armor plate patterning. So marine fish
covered from head to tail with armor plates, freshwater fish are lightly armored and fast,
they have greater body flexibility, usually only retain plates at the interior end. And
that dramatic difference in interior/posterior patterning of skeletal armor was used to give
different species names to marine and freshwater sticklebacks back in the 1800s. Same sort
of experiment: Here's a cross between fish that have a 30-fold difference in the number
of armored plates along this side of the body, and there's a single chromosome region that
controls three-quarters of the variation in armor plate number as well as several unlinked
We've also looked at some non-skeletal traits like pigmentation. Sticklebacks colonize water
that's either tea-stained, or clear, or different colors in different locations, so you can
get dark sticklebacks and light sticklebacks. If you cross the darks and the light ones,
there's a single chromosome region that controls 50 percent of the variation in pigment score
in particular body regions as well as unlinked modifiers. So what we're seeing, then, for
a range of evolutionary traits that have been selected in these populations is not a Mendelian
architecture; all of these traits you could call complex in the sense that they're controlled
by multiple chromosome regions. However, for armor, and for pelvis, and for pigment, the
genes of biggest effect control half or more of the variation in trait, and that's a very
nice result from geneticists' standpoint because you can explain a lot of the phenotype by
studying these major gene regions.
So we've spent a lot of time carrying out positional cloning projects to isolate the
genes for these other major chromosome region controlling armor and pigmentation. The armor
trait is controlled by a secreted singling molecule called ectodysplasin, actually named
after a series of clinical phenotypes identified in humans; the same developmental signaling
pathway is required for the formation of hair, and teeth, and sweat glands, and mutations
either in the signal. The receptor, the intersaler [spelled phonetically] adaptor molecule will
give that phenotype. In sticklebacks, the signal has been altered to produce this major
The pigmentation changes are actually due to a very well-known signaling molecule, one
of the most famous ones of all of Mammalian mouse development, called KIT ligand, or stem
cell factor. This is a secreted signal that's expressing in the skin, and the bone marrow,
and a variety of migratory paths, and the general ridge of mammals. The signal binds
to a receptor that's expressed on pigment cells, and blood cells, and germ cells, and
the signaling is required to direct the migration of those cells, to stimulate the proliferation
and their maintenance once they take up residence in different places of the body. So mice with
no mutations in KIT ligand are white. They're also dead. They die of absence of blood cells;
there's no germ cells that make it to the germ line. Another example of a sort of hopeless
monster from the standpoint of completely losing function, and yet that's exactly the
gene that has been selected in sticklebacks to produce pigmentation alterations in particular
So I think you see the similarities, then, in the themes that come from the studies of
pelvis, and plates, and pigment: In each case we get big phenotypic differences from a small
number of loci. The genes that have the biggest effects turn out to be these absolutely essential
key developmental control genes, and although those are required for the formation of lots
of different tissues, in each case we have evidence that there's been regulatory mutations
in the stickleback populations that has made it possible to produce the big advantageous
phenotype confined to a particular body region in the natural population.
And then finally, and I think very interestingly, we also, for every one of these traits, have
found that when the same thing evolves in different places around the world, exactly
the same gene is used, just like you've heard from Sarah Tishkoff for the lactase tolerance
gene. So I think that raises an interesting question about if it's used over and over
again in the different populations, how far the re-use of those evolutionary mechanisms
might actually extend. And so we've also been interested in trying to generalize from results
that we find in sticklebacks to characterize variation in other organisms, including humans.
So humans show all sorts of interesting traits, as you've heard about, including changes in
pigmentation. When we found that the KIT ligand is the basis of skin color changes in sticklebacks,
we also found that the KIT ligand gene shows one of the strongest signatures of selection
in different human populations around the world. So sticklebacks migrated out in the
ocean in the different water environments. Modern humans migrated out of Africa into
different environments. That migration is associated with very strong signatures of
selection in the non-coding regulatory regions of the KIT ligand gene, which we show were
associated with levels of skin pigmentation in the study a few years ago.
Let me show you what the selection signal looks like. So here's these big peaks of molecular
signatures of selection, centered not only on the coding region of the KIT ligand gene,
but the megabase of the non-coding DNA that flanks the gene, very high peaks seen in both
Europeans and Asians. So we've been very interested in trying to now track those changes down
to alterations in DNA sequence, identifying candidate regions based on molecular signatures
in humans, and then carrying out studies very similar to what I told you about in stickleback.
So here's surveys now of genomic fragments from the stem cell factor gene hooked up to
lacZ reporters instead of GFP reporters, and injected into mouse eggs instead of stickleback
eggs. But again, we're looking for the magic pieces of DNA that confer expression patterns
at locations related to the evolution of phenotype.
So let me show you what this looks like. Here's a serious of different constructs in the non-coding
regions of the KIT ligand stem cell factor gene. If you put those in the transgenic mice,
the left construct doesn't do anything, the right construct doesn't do anything; the one
in the middle drives expression in both skin and kidney. This is the fragment that contains
some evolutionary-conserved regions, and if you break it into two pieces, the left-hand
piece drives expression specifically in the kidney; the right-hand piece drives expression
specifically in the skin and the hair follicles, so even in this small region, two different
regulatory elements for expression in two different places in the body.
This one, obviously a great expression pattern for skin pigmentation. The endogenous KITLG
gene is known to be expressed in the keratinocytes of the epidermis and the hair follicle, hair-staining
endogenous KIT ligand antibody, and that's the expression pattern that we're recovering
in these lacZ constructs, with the small region of non-coding DNA upstream several hundred
kilobases from the KIT ligand transcript. Well, what's happened to that hair/skin enhancer
in different populations, there's a single base pair change that separates the sequence
seen in Africans and in Europeans. That single base pair change, when cloned upstream of
a luciferase reporter, doesn't eliminate the activity of the enhancer. It produces a small
quantitative change in the level of expression that we see.
Is that quantitative change actually a big enough alteration to produce a significant
phenotype? Well, to test that, again, when we're trying to look at whole animal phenotypes,
and this is challenging because you are trying to see what does one pair base do. The way
that we're testing that is, again, to hook those base pair changes not up to GFP reporters
but to the gene itself, either the ancestral or the derived hair enhancer that differ in
one SNP. We make transgenic animals using a special line of transgenics that has been
modified to have an insertion sight at a defined locus on chromosome 11. If we inject our two
versions of this construct together with an integrase, then these things will get integrated
at the same locus, same copy number, same orientation. So now you're scoring the two
alternative base pairs when it's the only base pair difference in the mouse. And when
we do that, here's the ancestral human SNP, here's the derived human SNP, and I hope you
can see this is a darker mouse than this one. That small base pair change is actually enough
to produce a significant difference both in hair color and the non-pigmented skin areas
of the mouse as well.
So pelvises and pigment are obviously very different, but I hope you can also see the
similarities between these two traits. In both cases, we've got an absolutely essential
developmental control gene that's surrounded by huge regions of non-coding DNA. Those non-coding
regions are chockfull, we think, of regulatory elements for driving expression at very particular
locations in the body, and in both cases, alterations have occurred within specific
enhancers: the deletion of the hindlimb enhancer in the Pitx1 gene; the single base pair mutation
in the hair enhancer of the KIT ligand gene that produced these dramatic phenotypic differences
that have been selected in unnatural populations. So key genes, but you can get viable, advantageous
phenotypes by making regulatory alterations. Many of the people in the room actually are
carrying a mutation, a regulatory mutation, in one of the most important signaling molecules
for Mammalian development.
Okay, last point I'd just like to end on is, I know these are case histories; we'd like
to know general patterns. And what this is -- one of the areas where the repeated nature
of stickleback evolution has been particularly useful, so because the same things have happened
over and over again, you can try to use genomics to recover all of the loci that are contributing
to the repeated evolution. So we published a reference genome of the three-spined stickleback,
and re-sequencing of 21 different stickleback populations around the world with the Broad
Institute last year, and the strategy that we used for looking for the adaptive loci
was to have repeated examples of the marine ancestral form and the freshwater form. We
line all those up with the reference genome, and then we window the genome, looking for
those places where all the marine fish look one way and all the freshwater fish look the
other way. So that's a site where the same thing is happening over and over again.
We were very interested to see when we just did that line-up and windowing of the genome-wide
sequencing data, you can make an index of how different are all the marine fish are
from the freshwater fish. You get this big peak here on chromosome 4. That sits right
on type of the ectodysplasin gene, which we had spent six years positionally cloning,
using high-resolution genetic mapping and transgenics to rescue the plates, et cetera.
You can recover exactly the same locus just by doing the whole genome sequence comparisons
of a whole bunch of marine and freshwater fish. You also see a whole series of other
peaks, both on this chromosome and on other chromosomes around the genome. In aggregate,
this has identified 84 regions. The median size of those regions is actually defined
pretty precisely because the breakpoints in the 21 different populations help narrow a
minimal interval that's being repeatedly selected, so it's a relatively small fraction of the
overall genome, but a very interesting fraction to us.
And so I'll end, then, by trying to use now not just a few examples but this larger set
to come back to an old question like whether evolution is controlled primarily by coding
or by regulatory changes. This has been a major debate in the evolutionary literature.
The actual answer from a whole series of these regions in sticklebacks suggests both mechanisms
contribute. We do recover, in this overall set of loci, a series that affect the coding
regions of proteins and produce consistent amino acid differences between marine and
freshwater fish. However, a large fraction of the regions map entirely outside of coding
regions in the non-coding intervals between genes, so those are regulatory alterations.
And then there's a fraction of regions that have both coding and non-coding intervals,
but the only changes that we see between the marine and the freshwater fish are in the
non-coding regions, and so, again, those are likely to be regulatory changes.
Very last point is that I think it's very interesting to compare these results to the
results that are coming from the kinds of studies that you heard about from Sarah Tishkoff,
looking for signatures of selection in different human populations around the world. We now
have a ton of data from HapMap and other re-sequencing studies that make it possible to look for
molecular signatures predicted to come along with a genomic region that has recently been
selected. Perta Sabetti [spelled phonetically] recently intersected a whole series of these
different signals to define 178 loci that show strong evidence for positive selection
during recent human evolution, and I went through this list and put them into the same
category that I showed you for the sticklebacks, and, again, there was a mixture of both coding
and non-coding regions, but I think what's really striking about the two different pie
charts is how similar the two distributions actually are. So these are incredibly different
scenarios: last 10,000 years, sticklebacks colonizing freshwater; humans 100,000 years
ago migrating from Africa to different populations around the world, and nonetheless, in both
of those scenarios, what you see is the overwhelming number of adaptive loci appear to be this
sort of regulatory alteration in genes.
So I'll end there, and just conclude by saying that I hope I've shown you that evolutionary
differences is at least a studyable, mapable problem. Major differences in natural populations
can come from relatively few changes. We see similar mechanisms being used over and over
again, and it looks like regulatory changes may provide the predominant bases for adaptive
evolution in both fish and humans.
Thanks very much. I'd like to end by saying thanks to a lot of people who participated
in the studies, and also major thanks to NHGRI, who made an investment in doing this sort
of comparative generics and comparative genomics work to try to study basic problems like the
molecular basis of producing new traits in natural populations.
Male Speaker: Thank you. So you've done a lot of great work
looking at the sequences that changes between all of these different populations. I'm very
curious if the networks between these different populations change as well, so if a mutation
in a certain gene actually creates new connections or loses other connections in the different
networks of these different populations.
David Kingsley: I don't think we have good data to answer
that. I can tell you that if you change a gene like Pitx1 which itself regulates a whole
series of other genes, then you have, in essence, changed a whole genetic program that normally,
if it turns on in this part body, will make a hindlimb. So yes, we do expect to see, by
the alterations here, a whole series of downstream events that actually are the black box that
connects the genotype to the phenotype, but we have not done all by all proteome interactions
and the sort of network analysis that you're also seeing starting to coming out in some
Eric Green: One more question there.
David Kingsley: Yeah.
Male Speaker: So, one theory for the loss of a structure
is that it has an adaptive value, a positive adaptive value. I was not quite sure why the
loss of these pectoral fins in the stickleback had a positive adaptive value.
David Kingsley: Yeah, so the pelvic reduction alleles have
been positively selected. So if you look at the signatures of the diversity surrounding
the mutations, that's an advantageous allele that's been swept in the population. The ecological
studies that have been done on the traits suggest that the loss of the pelvis occurs
in environments with low calcium and lots of insect predators when there are no predatory
fish, okay? So you need three interacting environmental variables: absence of a predator
that's retarded by the spine, the presence of a predator that uses the spine, and the
very low calcium levels, which also may create a metabolic cost for building the big bony
Eric Green: Okay, thank you, David. We -- just before
the break, we will have the first of several videos that we're going to share with you,
interviews and the like, so please enjoy the video.