Tip:
Highlight text to annotate it
X
[applause]
Elaine Ostrander: Hi, everybody. So thank you so much for coming.
So for the row of people sitting behind me, you do not get to fill this out until you've
heard me speak. So put your pens and pencils down there, all right? So how many -- how
many people in the audience have dogs? Yeah, so good this is a great, dog-friendly audience,
because we're going to be talking today about dogs, and a couple of different aspects: we're
going to talk about health as well as morphology, you know, why does are different shapes, and
how all that data sort of comes together to inform us about human health and human biology
as well. So we'll leave some time at the end; we'll have lots of time for questions. So
this is the closest ancestor to the dog. Who knows what we're looking at here?
Audience Response: Wolf.
Elaine Ostrander: What kind of wolf? Gray wolf, right, of course.
So dogs were believed to have domesticated from gray wolves. The latest data suggests
maybe around 13,000 years ago. So evolutionarily, that's really a -- that's a drop in the bucket.
I mean, that's not very long to go at all, and so one of the things we're always sort
of thinking about and pondering about when we look at our pets, wandering around our
house is, you know, is whatever they're doing caused by genes that are embedded in this
wolf genome, or is this something new that's come up during the process of domestication?
And that's one of the things that we go back to and we ask ourselves over and over and
over as we look at all of our friends.
So how many of you recognize your dog breed up there? All right, there's got to be some
golden retriever owners, I'm betting? There's probably some schnauzer owners, right? Maybe
some hound owners? You know, a Weimaraner owner or two? So these are a small number
of the 175 breeds of dog that are recognized by the American Kennel Club. The American
Kennel Club is the largest registering body of dogs in the United States today, but they
are not by any means the only registering body. There is the United Kennel Club, and
several other kennel clubs as well, and there are kennel clubs actually all over the world
that in aggregate recognize 493 different dog breeds today -- 493 different dog breeds
today. And those dog breeds show an extraordinary amount of variation, right? So we would say
that there's probably three things you need to keep in mind as we're going to talk about
dogs for the next 45 minutes, or so. It's important to remember that, even though all
of these guys look very, very different -- they have a different body size, different coat
colors, different head shapes, different leg lengths -- all dog breeds are members of the
same species. So they all have the same karyotype or the same chromosomal organization; they
all have the same genome organization; and they can be crossed to do produce fertile
offspring. Now, clubs like the American Kennel Club don't really encourage you to cross dogs
from one breed to the next, and as a matter of fact, to be, for instance, a registered
golden retriever, your parents had to be AKC-registered golden retrievers, and your grandparents in
turn had to be AKC-registered golden retrievers.
So one of the reasons people like me study dog breeds is because each one represents
sort of a closed population. So if you think about human populations that are isolated,
you know, people who live in Finland or Iceland, or people who have lived on islands for many,
many generations, or the Bedouin pedigrees, geneticists like to study those kinds of populations
because there isn't a lot of add mixture from the outside. They really have sort of a set
number of different alleles at each genetic loci, and that makes the problem of studying
complex traits, like diabetes, and cancer, and epilepsy actually an awfully lot easier.
The only problem in human genetics is that there's a limited number of such isolated
populations, whereas in the dog world we have 493 of them. And so we actually in my lab
have been working hard to get DNA samples from each one of those 493 breeds. We have
a great relationship with the American Kennel Club. We don't breed any dogs. We don't keep
any dogs in kennels, but if you go to a dog show or an obedience trial, or a specialty
or an agility trial, anywhere here in the tri-state area, you're probably going to find
someone from my lab there taking cheek swabs or handing out kits or collecting blood samples.
And we've now got a set of 50,000 DNA samples in my laboratory. So that's pretty impressive,
huh?
So when we look at these dog breeds, we really see extremes of variation, and one of the
things that makes a breed a breed is that that variation breeds true. So if you cross
one golden retriever to another golden retriever, the puppies are all going to pretty much look
the same, and there's some nice examples up here for you. You know, if you cross one Dalmatian
to another, they're all pretty much going to look the same. Cross one boxer to another,
they're all pretty much going to look the same. And the American Kennel Club is really
strict about the things that are important, and that define each breed. And it's different
for different breeds. For some breeds it's their coat color; for some, it's how tall
they are; for some, it's how long their legs are; for some, it's the shape of their face;
how far apart their eyes are; whether their ears are perked or whether they're down; and
when a dog goes into a dog show, those are all the things that the judges are actually
grading them on. So because breeders have been breeding for those traits for years and
years and years and years and years, those are the things that in my lab we're trying
to find the genes that control them, and in doing so, we're sort of getting a vocabulary
of growth regulation that we really can't get from studying worms, and flies, and mice,
and rats, and all those other traditional model organism systems.
Now, I really like this picture, and don't worry if you can't read the writing here,
but what we've done is we took 1,000 DNA samples from dogs representing 85 different breeds.
So we took about 12 different dogs from each breed, and we tested their genome at 100,000
different positions, and we were looking at the variation in the A's and those T's and
C's and G's that you're always hearing about. And then we fed it all into a computer program,
and we said, tell us how all the dog breeds relate one to another; and that's what this
wheel -- and you can really follow the color coding in the round -- is telling you so well.
So if you look at your upper right, you see there are the red, and you see spaniels, the
American cocker spaniel, the English cocker spaniel, the English springer spaniel, the
Cavalier King Charles spaniel, the Irish water spaniel, the Brittany spaniel, and so on.
And if you look over down at about 4:00, you'll see the sort of bull mastiff-type dogs -- the
miniature bull terrier, the French bulldog, the bulldog, the boxer, and so on.
Now, we have this hanging in my lab and we use this every single day as we're developing
a hypothesis. So if I'm going to study something like epilepsy, and I've got a whole bunch
of samples from English springer spaniels -- those are the red up there at about 2:00
-- then I can actually probably go out and get DNA from affected American cocker spaniels,
and Irish water spaniels, and Cavalier King Charles spaniels that are also affected, and
I can probably be correct when I hypothesize that they all got the disease because they
carry a mutation in the same gene. And that's because they all share a common set of founders;
that's why they're all colored in red. And I study epilepsy in the mastiff-like breeds
-- well, they probably have a mutation in a different gene, but when I look at the mastiff,
and the bull mastiff, or the bulldog, or the boxer, and the French bulldog, again, they're
all colored in kind of that teal color. They probably got the disease because, again, they
share a mutation in the same gene, and they likely got it from a founder.
So how long ago did this founder live? Well, how long ago do you think most dog breeds
were developed? You know, domestication occurred about 13,000 years ago. How long do you think
most dog breeds have been in existence? Any guesses?
Female Speaker: Two hundred.
Elaine Ostrander: You're pretty close. About 300 years. So most
dog breeds were developed in Europe, most were developed by fanciers during the Victorian
times. So most of them have only been around really for 200 or 300 hundred years. So again,
evolutionarily, we're not even talking about a drop in the bucket; we're talking about
a, you know, a perspiration drop in the bucket. I mean, as tiny as you can really get. So
we can make these kinds of hypothesis when it comes to a morphology or when it comes
to disease susceptibility, and we're usually right. And the other thing is that the genes
that we end up finding by studying dogs usually turn out to be important for the same things
in humans. It's just that, gosh, it's a lot easier to go to my freezer of 50,000 dog DNA
samples, and pick out a few related breeds, and look up their health records, and find
out what they've got and what they don't have, and correspond with their owners, and start
studying a pool of affected and a pool of unaffected dogs in order to find a disease
gene of interest.
Okay, so this is actually one of my favorite, favorite dog pictures. This is a Harlequin
Great Dane. It's skeletally among the largest of the dog breeds, and down here is a little
Chihuahua, which is skeletally among the smallest of the dog breeds. Now, we've been studying
dog morphology for several years in my lab, and we've published papers describing genes
that control body size, and leg length, and skull shape, and fur, and we rely on samples
from these extremes in the population in order to do studies on things like body size. Again,
remember, both of these guys are members of the same species, and they could be crossed
-- these two guys here -- to produce fertile if singularly unattractive, but nevertheless
fertile offspring, and you might want to give a little thought as to who would be the male
and who would be the female when you, you know. All right, so what do we have over here?
[laughter]
Elaine Ostrander: Okay, so this is, again, a Chihuahua; this
is Zeus. I had to put this picture in. Zeus actually holds the 2013 Guinness Book of World
Record as the tallest dog. So he is 44 inches at the withers, or the shoulders. This is
not photo shopped. This is actually how tall Zeus really is. And so we've been collecting
samples from dogs at these extremes, as well as everything in the middle, in order to fine
genes that are responsible for this dramatic, dramatic difference in body size that exists
within this one species. And I can tell you from a couple of papers that we've published
that there are about a half a dozen genes that account for most of that size variation.
And so I didn't put their names up, but if you're interested I can tell you they're the
insulin-like growth factor I gene; there are a couple genes on the X chromosome; there's
the insulin growth factor receptor; [unintelligible], HMGA 2, STT 2, growth hormone receptor. And
so some of these may sound familiar to you, because they've been shown in studies of mice
to be important in controlling mouse body size. The only thing is you don't learn a
lot from studying mice in this case, because you don't find mice that differ in size by
40 fold, right? I mean, wouldn't that be truly frightening, I mean, if you did, right? But
of course, we do find that in dogs.
So we've spent a lot of time identifying the genes and the mutations that account for variation
in body size. And we just published this, and I know it's kind of complicated, but I
wanted to show it to you, because I'm actually really excited about the implications of this
study. So if you look on the left, you see a bunch of bricks that are either yellow or
red, and I know it's probably hard to read on the bottom, but on the bottom is the name
of each one of these genes; one at a time: growth hormone, IGF 1 -- so on until we get
to IGF 1. And you don't have to worry about, you know, really which gene is which. Going
along the vertical are the weight in pounds of different dog breeds that we've assayed.
And then what the red and the yellow is telling you is that at each one of these genes, there's
two possible alleles, or there's two possible mutations or variants that can occur. One
is an ancient variant that we find in wolves, and one is a new variant that's only been
on the planet for a few hundred years, and we only ever see it in domestic dogs. Now,
the yellow is the ancient variant that we find in wolves, and red is the new variant
that we've only ever seen in dogs. Now, what's really striking is if we look at really little
dogs that weigh like zero to five pounds, all the way across you see red, telling us
that those dogs have the new variant, the one that we only see in dogs at every single
one of these genes, but as dogs get bigger and bigger and bigger, that drops off until
it's basically mustard yellow. And so all those big dogs are big because they carry
the DNA variant that came from the ancient wolf. So that's kind of cool.
So we took that data and we said, how predictive is it? You know, if I actually take what the
data's predicting versus what I really see in a panel of 500 dogs, how good is the predictive
value of my half a dozen genes? And that's what this line over here on you right is telling
you. And the fact that you get a pretty good line from all of those data points is telling
you just those six genes, if we could assay them in every single puppy that was born,
would have actually have great predicative value for what the final size of the dog is
going to be. We'd be right to about 82 percent. So think about this. This means that you could
go to the pound, you could get a puppy, you could get a cheek swab, you could get it assayed
at these six genes, and you would know what the final size of that puppy's going to be
pretty close to accurate, to about 82 percent accurate. So that's really amazing. Six genes,
just six genes control that much variation.
Now, one of the things about this study is it actually only holds true for dogs that
weigh up to 90 pounds, and if you think about the giant breeds -- the Great Danes, the Newfoundlands,
the Saint Bernards -- these six genes only have about 5 percent predictive value. So
there's probably lots of genes responsible for giantism [phonetic sp] in dogs that I
actually haven't found yet, and that's one of the things that my lab is in the process
of doing so that we can extend this line, and make it bigger and bigger and bigger until
we can actually understand all the genes controlling that full range of body size, going all the
way up to, you know, 180, 200 pounds.
Now, as we've begun to publish more and more of this data, people have gotten really excited,
and they've kind of picked out different things that they want to study, or that breeders
would like to try and breed for. So I have people in my lab studying leg length, and
studying skull shape; others are continuing to study body size, some are even studying
performance. And we built a big dataset based on 1,000 different dogs from 85 breeds, and
we also included 500 wild canids -- so coyotes, wolves -- from all over the world. And we
tested their DNA at 1,000 different points in the genome, and we made that DNA publicly
available without any restrictions, without any patents to anyone who wants it. So you
know, we really encourage other people to try and think about these problems as well.
Now, I like this particular story. This is one that was led by Heidi Parker [phonetic
sp], and she was a graduate student in my lab, and then she went on to be post-doc,
and then she went on to be a staff scientist, and she's been with me for 15 years, and I
actually don't think she's ever going to leave. So Heidi has been really interested in short-legged
dogs. And there's about 20 such breeds. They're called chondrodysplastic. They have a ratio
of height to body length less than zero. So they sort of have a normal head, and a normally
proportioned body, but then they have these very short and thick legs. In the dog world,
we would say their structure is well-*** or heavy, and their forelimbs, like you can
see on that basset hound, are often sort of bowed, or a little bit curved out.
So Heidi came to me one day, and she said "I want to try and find the genes responsible
for this trait across these 20 breeds." And I said, well, you know, this could be a really
hard problem. It could be that they all share a common mutation, but these breeds were developed
for different purposes. Some were developed to be -- to go down rabbit holes; some are
fox hunters; some are companions; some are ratters. So I said I'm not so sure they're
all going to have the same mutation in the same gene. And she said, "Well, I think this
is a really interesting problem; I'm going to try it anyway." And she's not being totally
unbiased here, because you see that doxen and that basset hound? Those are actually
her dogs.
[laughter]
Elaine Ostrander: Yeah. And so, you know, when you come to my
lab and you work, people are often motivated by things that they have with their own dogs,
and you know, some of my graduate students are top -- have top show dogs, or top dogs
in the agility ring. So it's not -- I've actually had mushers come and be graduate students
in my lab. So, again, motivated by trying to understand the underlying genetics.
So Heidi went after the gene for chondrodysplasia. Now, I know this is -- looks like just a bunch
of lines to you, but what you're looking at here in the alternating gray and black, each
one represents a different dog chromosome; and dogs have 38 chromosomes, and we also
included the X, but we didn't put in the Y, because we figured there was nothing important
on the Y anyway. And then what we're doing is we're comparing cases, who are chondrodysplastic,
to what we call control, which are all the other breeds that are not chondrodysplastic;
and we're looking at 100,000 different points in the genome, and we're saying, is there
a chromosome that has a data point that's significantly different between cases and
controls? And you can see this is very significantly different here on canine chromosome 11. And
for those of you who've calculate P-values, the P-value here 10 to the negative 102. So
.0000000 -- put 102 O's, and then a one, and that's how statistically significantly this
is. So this is hugely statistically significant.
So we decided to follow up on this, and the way we did it is we looked for something that
evolutionary biologists call a selective sweep. So what's a selective sweep? Well, when you
have a selective sweep, you assume, as I've told you before, that there's an ancestral
mutation that occurred many, many generations ago. In this case, before dogs were divided
up into lots of different breeds. And then dog breeders breed, and they breed, and they
select for different things, and there's a lot of scrambling of chromosomes, but the
mutation stays, because they're always still selecting for that one trait; in our case,
chondrodysplasia. And so now, when we look at modern day dogs, their chromosome may look
nothing like the ancient chromosome, except for where the mutation is, and in the space
right around the mutation. And so we look for that region of commonality, and when we
find a region of commonality across a group of breeds who have a trait, then we know the
mutation has to be somewhere in there, and that turns out to be exactly the case here.
So this is an old fashioned gel, and I put up because I think probably some of you in
college had a -- have had a chance in science classes to run a gel. A gel simply separates
DNA based on its size. The control dogs are things like greyhounds, and boxers, and cocker
spaniels, and if you look at the top, you don't really see anything of the gel. But
when you look at the case dogs, and these of course basset hounds, and the doxens, and
the Pekingese, you see a bright yellow -- a bright white band. And that's telling you
that all of these cases have some extra DNA that's responsible for this trait that's not
present in the group of controls. So right away we know that what our mutation is -- it's
not a single base pair change, and it's not a loss of DNA. All these chondrodysplastic
breeds have acquired extra DNA. And in fact, they've acquired an extra copy of a gene called
fibroblast growth factor IV.
Now, they didn't acquire any of the regulatory machinery that tells it when to turn on or
when to turn off, but they acquired the full sequence of the gene. And so actually the
genes around it, they sort of parasitize the regulatory machinery from genes around this
-- what we call retro gene, and that's telling it to be expressed in fetal chondrocytes,
and you know what that's doing? That's closing the growth plates prematurely. So the legs
never elongate as long as they should. This gene is expressed, the growth plate closes,
and er, er, er, er, er [phonetic sp] -- the leg can't elongate to its full and natural
length. And every single one of those 20 breeds I showed you has exactly that gene, including
this, which is the corgi breed.
So this was really exciting, and we were able to publish this 'Science,' not because the
editors of 'Science' care about corgis, although I think they should, but because this was
sort of a new way to screw up the genome that had never really been described before in
mammals. And the other thing is, we of course know that there are humans who suffer from
forms of what we've historically called dwarfism, but it's really chondrodysplasia, and we don't
always know what's causing that. So now this gene goes into that lexicon, into that vocabulary
as something that we need to think about when we look at those. So this is a really neat
example, just by studying a phenotype that dog breeders have been breeding for for a
couple of hundred years. In a bunch of healthy dogs using DNA from my freezer, we've been
able to figure out a whole new mechanism for screwing up the genome, and we've been able
to add a gene to the medical genetics vocabulary that turns out to then become very important.
So this to us is really a huge success.
So we've gone on to do that in several other ways, and I'm going to give you one more example
in morphology, and then I'll give you one example in disease, and then we'll have some
time for questions in the end. What are these?
Audience Response: [inaudible]
Elaine Ostrander: Absolutely. These are all skulls from dogs,
and these are pictures we took down here at the Smithsonian. Turns out they have a lot
of skulls in the back room.
[laughter]
Elaine Ostrander: Right. And these are all different dog breeds,
and they differ in both shape, and what else?
Audience Response: Size.
Elaine Ostrander: And size, exactly. So when Jeff Shaunenbeck
[phonetic sp] joined my lab, he said "I want to find the genes that control this. I want
to understand the genetics of the skull shape and size in dogs." And I said, Well, I don't
know. This sounds like a really hard problem, but Jeff had one of these giant Leonberger
dogs with a big round, kind of fluffy face, and so, you know, this is what he wanted to
do, and so that's what we did. Now, the first problem we had was we don't know how -- wow
do you quantitate a skull, right? I mean chondrodysplastic is easy. The dogs either have it or they don't
have it; and body size is easy -- you measure them or you weigh them. How do you actually
quantitate a skull? It turned out to be a hard problem, but we solved it, and what we
do is we have something called a MicroScribe Digitizer, and we touch each skull at 51 different
landmarks, and that sends data to a computer, and then in the end, the computer takes the
51 data pieces and it draws a three dimensional picture of what that particular skull looks
like.
So here I'm showing you top and bottom and side views of a particular dog's skull, and
everywhere there's a red or a blue number, that's one of those data points that we've
gotten. So the palate can be long or short -- that's the roof of your mouth -- for a
dog that has a long or a short snout; that angle in picture C between the rostrum and
the nose, that can be, you know, pretty much a ski slope, or it can be at a right angle,
like it is in a Newfoundland. We would say they have a Roman nose, or a very high forehead;
and there's variation actually in every one of these traits. Now, we've actually been
fortunate to travel around the world. We've measured about 1,000 skulls from 161 different
breeds, and people always ask me, where do you go to do this? And we go to lots of museums,
and certainly the Smithsonian is the first place we went, but we've been to museums and
universities all over the country, and actually all over the world. The university in Switzerland
actually has something like 2,000 canine skulls that we're about halfway through measuring
now.
But I have to admit, there are a group of people in the United States that have a lot
of skulls in their basement, and I don't why, but they do. And they all call me, and they
say, you know, I got a bunch of dog skulls. Why don't you come down and measure them?
This is Alex, one of the people in my lab, and she's measuring the dog skulls. This is
a gentleman in California, who called us to fly out and go down to his basement, and measure
all the skulls, and you can see he has all kinds of animal skulls. We really don't know
why. We don't ask those questions. We just measure the dog skulls, and get out of there,
but yeah, there's a lot of these people in America. They all want to friend me on Facebook.
You know, it's a whole culture thing. So anyway, but these people have been very generous with
their collections, and we've actually gotten a lot of data from skulls where we could verify
the breed and we could verify the age.
Okay, so this is -- this is in some way sort of a tragic slide. So this is really what
motivated Jeff to begin this project. So in the left-hand column are a set of human conditions
that are really different kinds of craniofacial abnormalities, and I know not -- you're not
familiar with most of those words, but they describe different abnormalities that we see
unfortunately in humans, often associated with particular syndromes. And next to them
are breeds where the breeders are actually trying to breed that in as part of the breed
standard, and really one of the examples that I use -- and this is not something that the
American Kennel Club is doing, this is not an AKC breed, but this dog shown here is a
Pachón Navarro, and the breed standard includes having this deep cleft that goes all the way
from the outside of the snout, all the way down in to the roof of the mouth. So you know,
this is not something that American breeders are advocating, but you know, it is something
that, you know, we see and so we want to try and find those genes, because we think that'll
help us understand something about cleft lip and cleft palate, and we're actually interested
in all of these craniofacial features.
So the one that we've been doing the most work with is called brachycephaly, and that
means having a very pushed-in face. So if you think about your Saturday morning cartoons,
those of you that are old enough, and some -- you know, something happens and, you know,
the face just kind of accordions in like gnaaa [phonetic sp] -- kind of like that, right?
And so that's the kind of appearance you see in the pug or the cane corso, and that's very
different than what you see in the Afghan and the bull terrier, which have very elongated
noses, and that's referred to as a dolichocephalic phenotype. So a long nose versus that very
pushed-in face. So we've been going after these kinds of genes, and we do the exact
same experiment over and over, we comb the genome, and we look for evidence of a selective
sweep, and that's what the data actually looks like. So these are base pair positions along
canine chromosome 30, and you see a certain level of chatter, and then when you get right
here, you can see this big dip. And that's telling us that there's a selective sweep
there, that that's a place where there's a lot of homogeneity, that -- something breeders
have been selecting on for years and years and years and years. So they don't know that
there's a gene under there. They don't know what the gene is they've been selecting on,
but we're going to find it, and we're going to tell them what it is.
Okay, so here are a set of 12 dog breeds that my lab has now sequenced. We've sequenced
these breeds pretty deeply so we've got a pretty good genomic sequence, and we picked
these breeds for lots and lots of reasons. We have lots of different studies going on,
and there's another actually 40 dog breeds that we and others have sequenced as well,
but these were the first 12. And partly, we picked them because if you think about going
very brachycephalic to very dolichocephalic, you have a really nice continuum. So we will
be able to use this data to try and figure out what's underneath that little V, that
statistical blip that the breeders have been selecting on over and over by comparing the
brachy- and dolichocephalic dogs.
Now, you've heard a lot about the genome project, and usually people talk about the human genome
project. When I hear about the genome project, I think about the canine genome project, because
that's what's really important to me. So I know this, again, looks like a checkerboard,
but each one of the rows is giving us information about the sequence variation we saw in those
12 dog breeds. So we started out with 190,000 base pairs, and 2,000 possible variants, and
then we started filtering and filtering. We got it down to 85,000 variants and 48 variants,
and in the end, when we applied the sequence of all those other breeds, we were able to
find the single base pair in the single gene that turns out to be important, and it's how
canine chromosome 30 contributes to that facial phenotype. So the gene is called bone morphogenesis
protein 3 -- kind of makes sense it's a bone morphogenesis protein -- and it is a single
base pair change that changes one amino acid, phenylalanine to lucy [phonetic sp]. One amino
acid.
Now, when Jeff came to me with that data, I said, gosh, you know, I believe you, but
in order to publish this, we're, you know, we're going to have to have more proof. And
so Jeff did lots of statistical studies, and they all looked really, really good, and we
wrote it up, and I said, you know, it's pretty good, but I think we're just going to need
just a little -- a little bit more proof to get it into one of the fancy journals. And
Jeff said, "Well, okay. I can knock that gene out in zebra fish, and I can make a pug-nose
fish." And I said, "You're going to make a pug-nose zebra fish?" And he said, "I'm going
to make you a pug-nose zebra fish." And so there's a technique he used, and he knocked
that gene out, and that is exactly what Jeff made.
So let's forget the top row for a second -- I'll come back to that -- but if you look where
it says E, F, and G, that's a zebra fish, you know, several -- about 48 hours after
the embryo was fertilized, and it didn't get injected with the stuff that knocked out the
gene. And then the blue is the cartilage from the top of the jaw, the top jaw, and the G
is the bottom jaw. And now, the other two are examples of fish that did get injected
with what we call morpholinos, and they knock that gene out, and you can see that the jaw
is gone, especially that bottom jaw. Look, there's almost no blue staining in J and M,
and even the top jaw is pretty screwed up as well. So in essence, Jeff made me a pug-nose
zebra fish, and indeed, in doing so demonstrated that by knocking out just that one gene, we're
able to dramatically affect the jaw structure. And so this gene, in fact, does turn out to
be important in one particular type of human craniofacial abnormality.
So this is another example of how, you know, we started with healthy dogs, long-lived dogs,
but they had a particular phenotype, this sort of pushed in face, and we've been finding
the genes responsible for that. And there's not just one; it turns out there's actually
several. And in aggregate, they are responsible for the dramatic difference between being
brachycephalic with a pushed in nose, or dolichocephalic with that elongated nose. And we can prove
we're right by going to some of these model organisms like zebra fish, which are these
sort of little tiny fish that people use in the lab sometimes.
Okay, so those are examples of how my lab has been studying morphology. And right about
this point in time, someone usually says to me, well, you know, that's great, but -- gosh,
dogs have an awful lot of diseases. Do you actually, directly study any of those diseases
as well, and if you do, are they telling us something about human disease? And the answer
to both those questions is a resounding yes. So here are the top 10 genetic diseases in
dogs, and what do you notice is number one?
Audience Response: Cancer.
Elaine Ostrander: Right. About one in how many humans will get
cancer in their lifetime? Anybody? About one in four people will get cancer, some kind
of cancer at some point in their life. They may not die of it, but they'll get it, and
about one in three dogs will get cancer in their lifetime as well. How many of you have
had a dog or a cat who got cancer? Yeah, right. So in my lab we do in fact study several different
types of cancer. Now, this is one of my favorite dog pictures ever. It was actually sent to
me by a very well-known, and very generous dog breeder, and she breeds -- what are these?
Audience Response: Poodles.
Elaine Ostrander: Standard poodles. Right, these are standard
poodles. So about four years ago, we started getting phone calls from people who owned
standard poodles, telling us that the dogs were getting a particular kind of cancer,
and it's called squamous cell carcinoma. And oddly, it was occurring in the toes, and sadly
the way the veterinarians have to treat it is they have to actually remove the toes.
And so that's horrible for the dogs, it's horrible for the owners. If the dog's a show
dog, it won't be after that, and people obviously don't want to breed to those dogs after that,
and so this is really a big deal for this community, but what was so interesting about
this is they said, you know, we only ever see the disease in black standard poodles,
and we never see it in white standard poodles. And so we've now looked at hundreds of standard
poodles, and we see it in black and brown, very, very, very, very, dark gray, but we
don't see it in the white, or the cream, or the apricot dogs. So we thought, well, you
know, there's a lot of selection for coat color in standard poodles. Maybe they've inadvertently
selecting for a cancer gene as well, and that in fact turns out to be the case.
Now, let me just tell you that we study lots of kinds of cancer in my lab, and in fact,
across the world dog geneticists study lots of kind of cancer. So those of you who have
long-limbed breeds like Scottish deer hounds or Irish wolf hounds probably worry about
osteosarcoma. We see tons of bladder cancer in Scotties, and Westies, and Shelties, and
we actually have a paper we're writing about that now that comes again from sequencing
tumors, the DNA from tumors that we find in those dogs. If you've got a Bernese mountain
dog, a super wonderful dog breed that's really increasing in popularity, or a flat-coated
retriever, one in four, one in six of those dogs will get malignant histiocytosis or histiocytic
sarcoma. Stomach cancer, we see in the Belgian sheepdog, the Belgian Tervuren, as well as
in the Chow Chow. Universally lethal; dogs don't survive it. And the idea here is we
study these in dogs because the breed structure, as I told you at the beginning, simplifies
the overall problem. We'll see shared genetics among affected dogs, and it'll be distinct
from what we see of healthy dogs of the same or, remembering the wheel, very related dog
breeds.
So this is a picture of squamous cell carcinoma. You can see that the toe is sort of blown
out. It's the most common nail bed cancer in dogs. If you are a giant schnauzer, your
chances of getting this are 22-fold higher than the average mixed breed dog walking down
the street. If you are a Briard, your chances are 10-fold higher than the average mixed
breed dog walking down the street, and if you're a standard poodle, they're about 6-fold
higher, although standard poodles are where we see most of this, because they're the most
popular of those three breeds. And again, in the cases of the standard poodles, we only
see it in the black dogs, not in the white. For really complicated reasons in the Briard,
we actually see it in black as well as white dogs, and we figured out why that is. It's
actually a very, very complicated genetic story, but let me tell you a little bit about
standard poodles.
So we, again, combed the genome with our 100,000 points of variation, and we found a signal
on canine chromosome 15. It wasn't as strong as what we saw when we were looking at those
morphologic traits, because breeders haven't been trying to breed cancer into dogs the
way their trying to breed, you know, short legs or large body size; but nevertheless,
it's there and it's in a lot of dogs. Now, I think this is the last data slides that
I'll show you. And so what we did is we found a region on canine chromosome 15. It was about
a million base pairs long. We sequenced it in lots of affected and non-affected dogs,
and everywhere that there was a possible mutation, there is a triangle. And then, we defined
a region that, you know, maybe if we were liberal in our thinking, was about 500,000
bases in size; and if we're conservative in our thinking, it's about 800,000 bases in
size. And what's cool is there's only one gene in that region, and that gene is called
KIT ligand, and we knew instantly that we had found the right gene, because it's a gene
that's important in coat color, but it's also a gene that has been shown to be important
in cancer. So we did a lot of work. It took about three years, and we in fact, found the
mutation, and in this case, we again found sort of a new and interesting way that the
genome gets screwed up.
So it turns out that the mutation is, again, extra DNA; it's not a deletion. It's an insertion
of about 5,000 bases, and it can be present one, two, three, four, five, or six times.
And the more times it is present, the more these green proteins bind; and the more these
green proteins bind the more they ramp up production of this gene KIT ligand; and the
higher you ramp up production of this, the greater the chances are that you're going
to get the cancer. So if you're a dog who has this insert present on both of your chromosomes
four or more times, boy, your chances of getting cancer are really, really high. If you have
it maybe four times on one chromosome, three on the other, it's sort of moderate. But four
is really the threshold. If both of your chromosomes have it repeated three times or two times
or one time, or any combination of thereof, no chance you're going to get cancer. No chance.
And we've looked at hundreds and hundreds of dogs, and we've never even found one. So
this is one of these sort of threshold deals where you have to get KIT ligand ramped up
to a certain point in order for it to go ahead and cause the cancer. So it makes it hard
to develop a genetic predictive test, but people are in the process of doing that.
We found out why white standard poodles didn't get this. It isn't because they didn't carry
the mutation, the four-four genotype. They didn't -- they do indeed have chromosomes
where this insert is present multiple times, but they have a compensatory mutation on another
gene called MC1R, and it completely knocks this out. So it's a case where they have a
bad mutation, but they have a good mutation and good triumphs over evil in this case,
and so they never, ever get this, even though they carry the bad genotype. So it's an important
lesson, because just looking for the presence or absence of the bad genotype isn't really
totally predictive. You also have to look for the presence or absence of the compensatory
mutation, and all the white standard poodles have it, none of the black standard poodles
have it. And so that explains the difference in what we observed in coat color.
So I'm going to stop there. We have about 15 minutes to ask questions, or 12 minutes
or so. I hope I've showed you that dogs are a really fun system for looking at both simple
and complex traits, including susceptibility to cancer, which is of course a very complex
trait. When we study morphology in dogs, we learn things that are important about development
of all mammals, and that includes humans as well. For both canine and human health, these
studies of cancer have become very, very important, and there are labs really at vet schools,
as well as at non-vet schools that are studying every conceivable kind of cancer, as well
as all the other common human diseases: epilepsy, diabetes, heart disease -- you know, whatever
you've got including morphologic traits you don't like, like baldness or obesity or things
like that. Those -- there are labs all over the world that are studying those things as
well.
Many of these studies are long term; some are short term. There are a number of disease
genes that we've been able to find the mutation for. We've been able to develop a genetic
test, and breeders are now using those genetic tests to make really good decisions to produce
healthier, more long-lived dogs. So you don't see kidney cancer in the German shepherd anymore,
and we've been able to wipe out collie eye anomaly, which is a degenerative disease of
collies and border collies, and a number of herding breeds, as well as several other diseases.
And this has all been done in collaboration with breeders and owners and veterinarians
actually all over the world.
Samples are always needed so if any of you have a really interesting dog breed -- I gave
this talk yesterday, and a man came up to me afterwards and said "I have a Shiba Inu
if you want DNA from that." Yes, we do want DNA if you've got an interesting dog breed.
We love your labs, we love your golden retrievers, we love your German shepherds, but we probably
have enough DNA on those, but the more esoteric breeds, we're still always collecting more
DNA from. And you know, great progress can be made, but you know, it is necessary to
get the DNA samples, and allow us the time to do our work. So for the breeders in the
audience, I know many of you have contributed samples, and you wonder how long it's going
to take. Wel, sometimes it's a year, sometimes it's going to be five or six years, because
some of these problems end up being simple; some of them end up being very complex, but
our goal is always to make available to you some sort of a diagnostic test.
So I'm going to go ahead and stop there, and allow you to -- turn up -- maybe turn up the
lights? Turn up the lights, and I'll go ahead and take questions. Okay?
[applause]
Elaine Ostrander: Fire away.
Female Speaker: [inaudible]
Elaine Ostrander: This slide? How do we find out what?
Female Speaker: How do you find out the disease risk?
Elaine Ostrander: How do we find out the disease risk? So what
we did is we got DNA from a lot of poodles who had the disease and a lot of poodles who
didn't have the disease, and then we sequenced it to see how many times this was re-iterated,
how many times this particular 5.7 base pair unit was repeated. And we actually knew from
looking at the human genome, we knew exactly what these 5,000 base pairs do. We knew that
this protein binds to them, and when it does it ramps up production of KIT ligand. So we
can do statistics looking at dogs who have two copies, three copies, four copies, five
copies, and we can look at them at 10 years, 11 years, 12 years, 13 years, 14 years, and
see who does and doesn't have cancer. And from that, we can figure out what the risk
is for a dog who has three copies or four copies or five copies, and four is really
the cutoff. If both of your chromosomes have three copies or two copies, or one copy, we've
looked at hundreds of dogs, and not a one of those dogs has this kind of cancer, but
as soon as one chromosome gets four cancer -- has the repeat -- repeated four times,
then we start to see the incidents of cancer creeping up, and the more you have, the higher
it gets. Okay? And there are you know formal statistical tests that we can apply to that,
and if you're interested, if sort of got a statistical mind, come see me afterwards and
I'll give you the paper and the tests that we use. Okay, other questions? Yep?
Female Speaker: I notice that you were [inaudible] dogs [inaudible]?
Elaine Ostrander: Sure. So the question was, is epilepsy indeed
prominent in dogs, and yes, we see it in lots and lots of dog breeds; and have we found
genes? We've been part of one study that identified one gene, published it with a group in Belgium
several years ago in Science, but lots of other people are looking at this problem,
because you know, not only is it an important problem in human, it's a big deal in dogs.
I mean, if your dog has epilepsy, I mean, that is a lifelong problem that you as a pet
owner have to deal with, and certainly those dogs don't show and people don't, you know,
put them in the breeding pool if they know. And so some of those genes have been found;
not all of them have been found. And there are some breeds where it is a much worse problem
than it is in other breeds. So that's one of the areas where we see some of the most
active and intense work. Yeah, you had a question.
Male Speaker: This is sort of a general [unintelligible]
Elaine Ostrander: Okay, I probably won't know the answer, but
go ahead.
Male Speaker: Well, I was just wondering what sort of breeds
made up the dog reference gene?
Elaine Ostrander: Oh yeah, sure I do, because I picked the breed
for the reference genome. So he asked me, what breed or breeds make up the dog reference
genome? So it's one breed, and it's one single dog. So in 2001 I had the chance to pick the
dog that was going to be sequenced, and you know, truthfully what I did is I probably
looked at 100 different dogs of all breeds, and I picked the most in-breed dog I could
find. And the reason for that is the way genomes are sequenced is not from the top of chromosome
1 to the bottom of chromosome 38. What they do is they cut the DNA up into a zillion little
pieces, and then they sequence it all randomly, and then they have a computer put it back
together. So if what you got from mom is really different than what you got from dad, it's
a harder computational problem. If what you got from mom and dad was pretty similar, it
a much simpler computational problem.
So I tested 100 dogs at 100 places in their genome, and there was one dog that was easily
the most in-breed. It was a boxer, and it was from New York state, and as luck would
have it, it was actually owned by a veterinarian. So when I went to him and I explained that
his dog was the lucky winner -- I mean, he actually understood that this was really important.
He provided us with an awful lot of DNA, and he understood that, you know, when the dog
-- it was a pet, you know, like a lot of our dogs are -- that when the dog died that we,
you know, we would like to get some samples from that dog, because it was, in fact, going
to be the reference genome. And his only request that I not divulge his name, which I've never
done, or his location, which I've never done. So he was great, he was fantastic, but it
was a boxer, and her -- it was a she. Almost all the reference genomes, except maybe maybe
-- are a she, because then we get good data on the X chromosome.
Female Speaker: Have you guys found a gene related to like
a Yorkie, just randomly like [inuadible]?
Elaine Ostrander: So is -- have we found a gene linked to fur
loss in breeds like the Yorkie? So I haven't looked at that. There are breeds, like the
Chinese crested or the Mexican hairless, there are breeds that have very little fur, except
for some tufts at the top of their head and down by their toes, and those genes have in
fact been found, but those dogs, that's part of the breed standard; that's how they're
supposed to look. When a dog blows all of its fur, I mean -- and that's an anomalous
sort of thing -- I'm sure there are people looking at it. I'm not one of them, and I'm
not aware of a paper, but I could certainly tell where in the literature to look for it.
Yeah?
Male Speaker: Are you able to define what would be the top,
like healthiest breeds?
Elaine Ostrander: Sure. Yeah, so the question is what are the
healthiest breeds, and it's a little bit of a trick question, and the reason is because
within every breed there are lines that are really healthy, the breeders are very savvy,
they've gotten dogs from multiple places in the world that are members of that breed,
or multiple places in the United States that are members of that breed, and they've worked
really hard to maintain the hybrid rigor of that breed; and there are other breeders,
you know, who have bred a lot of closely-related individuals one to another, and their lines
may look good, but they have a lot of health problems. So there's no one right answer.
I mean, you can look on the internet and, you know, they'll certainly give you those
-- that kind of information, and they'll say things like, well, you see a lot of cancer
in boxers or in golden retrievers; or you see a lot of copper toxicosis in the Bedlington
terrier. You see -- you know, there's different diseases that tend to predominate in different
breeds. There are some situations where disease is so predominate, like copper toxicosis was
in the Bedlington, before the gene was found that I think that was a fair comment. You
know, there are other cases where I think you just have to really search. I mean, I
wouldn't be hesitant to own a golden retriever or you know, a German shepherd, or any of
these other really popular breeds. I would just work really hard to talk to people the
breeder had sold to, and find dogs that lived a really long -- a really long life.
In general, small dogs live longer for sure -- the terriers, the toy dogs -- in general,
they do live longer. And dogs that have a working job like border collies, boy, some
of those live 17, 18 years. My border collie lived to be 13 years. So breeds where there's
a working lineage -- I mean, they have to healthy to work -- those tend to be pretty
long-lived. The really big breeds tend to have more heart difficulties and problems
-- Saint Bernards, and you know, some of these giant breeds where they've just almost been
bred to be too big for their heart. You know, they rarely live beyond the age of 10 or 11,
because they almost always go for -- from heart problems. So, you know, I always tell
people to, you know, pick the breed you want, pick the breed that matches your family. There's
all kinds of tests on the internet to help you do that. But then work with -- you know,
really look around and take your time to find a healthy and reliable breeder. Yeah?
Female Speaker: Would you say that breeds that are closer
to the wolf, like in ancestral [inaudible] lineage, would you say that they are [inaudible]?
Elaine Ostrander: No. So she -- the question she asked is, are
breeds that are more related to wolf healthier, and I guess I would say no, you know, in part
because things that maybe look more wolfish, like the Malamute or the husky, I mean, you
know, there were clearly multiple domestication events that occurred in multiple places in
the world, and you know, by and large, truly it's the small dogs that are the healthiest
and really do live the longest, and so I wouldn't necessarily say those are, you know, among
the healthiest, although, you know, I had a Malamute that lived to be like 15 or 16
so, you know, again dogs that have a working function, but this was bought from a line
of dogs that, you know, were involved in mushing, and so that's -- you know, those are going
to be healthy dogs so I wouldn't necessarily say that. Other questions? Yeah?
Female Speaker: Do mixed breeds do better than pure breeds?
Elaine Ostrander: Do mixed breeds do better than pure breeds?
You know, everybody thinks that they're going to, and I have so many people come up to me
and say "I bought a Labradoodle or a" -- you know, they have all these weird, cool names,
and they said, you know -- and so -- hybrid, bigger, you know, I'm -- my dog's going to
live to be 22, and then they're shocked when they don't. And the reason is because, you
know, if the parent breeds or the parent lines were themselves not healthy, particularly
if they were not healthy because they had the same disease or the same mutation, you
haven't -- it doesn't do any good, right? And so you know golden retrievers, you know,
they get a bad reputation for having a lot of inherited disease. I think there are some
lines of golden retrievers that are wonderful out there, but there are also some lines that
really do have a lot of disease, and because they're so good with families, they're involved
in an awful lot of crosses, and you know, you can say that of lots and lots of other
breeds. So, you know, in general, what I like about pure bred dogs is you sort of know what
you're getting. You know, most poodles I know live long, long, long, long periods of time.
There an awesome breed in all three sizes. They have a reputation in the breed club for
being very vigilant, being very careful, and so I don't necessarily think that's really
the best choice. Yeah?
Female Speaker: [unintelligible] you talking about being able
to remove the mutation? So is that happening widely within that breed --
Elaine Ostrander: So --
Female Speaker: -- or is there some special [unintelligible]?
Elain Ostrander: -- the question is, you know, removing mutations,
does that happen widely? So it really isn't removing mutations in the sense of how that
language is used, but it's really about breeding it out. So breeding a carrier not to another
carrier, but to a healthy dog. And when I talk to breed clubs, I tell them first thing,
don't throw all the carriers out of the breeding pool, because they're contributing so many
good things that you're going to just end up with something else you didn't have a problem
with before. But -- but take your carrier and breed him to a non-carrier. The carrier
-- get the progeny tested, breed carriers to non-carriers, and gradually breed it out.
And we've seen examples where a small breed has thrown out all the carriers, and they
have a disease that's really prevalent, and then suddenly they have three other diseases
that crop up as recessive.
The American Kennel Club has just been wonderful. I mean, they ask us to come to all their meetings
and their specialty events, and their trials, and we are inundated with invitations to come
and give talks about exactly these kinds of issues, and labs like mine often make the
data available without patenting it. Sometimes it is patented, but often without patenting
it, just so the tests can get out there and people can start using it. And breeders are
some of the smartest geneticists in the world. They absolutely use it, because it is their
livelihood. They want to breed healthier, more long-lived dogs, and if they're the first
off the block with a reputation for using genomics to breed healthier more long-lived
dogs, people love that -- people love that. And so I have found that this -- I mean, I
have 50,000 DNA samples in my freezer, and I can think of one incidence where I asked
for a DNA sample and I was turned down. Go ahead.
Male Speaker: My question is back to the mixing question.
So in agriculture, the way you find new traits [unintelligible] is by crossing breeds so
you will get the non-additive effects and things like that. Is there a push in the dog
world to try these things out and see what happens by mixing different breeds?
Elaine Ostrander: No. No. There isn't.
Male Speaker: [unintelligible]?
Elaine Ostrander: Right. So the question that was asked was,
is there a push in the dog world to, you know, mix dogs of different breeds to try and figure
out what's going on with the genetics of some of these traits? I just put this up because
it's my favorite picture in the slide show. And the answer to that is no. First of all,
we don't breed or keep any dogs. So we're not going to do that. And in the pure bred
dog world, you know, people -- you know the convention is you breed dogs of one breed
only to members of the same breed, and that's just not the convention in that community,
you know, to take a dog that may be a popular sire and breed him to a bunch of other breeds
to figure out what's going on. That's just not what is done. So it really isn't. And
you know, I'm not in the management of the American Kennel Club so it's not really for
me to say, but -- okay, one more? Can take one more question? All right. One more question.
Sure.
Male Speaker: You mentioned about 50,000 data points. Are
they all pure breed or do you also have mixed breeds and mutts?
Elaine Ostrander: So I take 100,000 data points, and I have
50,000 DNA samples from dogs in my freezer. I have -- my lab has very few mixed breed
dogs. Almost all of those are from pure bred dogs. They're not all American Kennel Club
recognized breeds. You know, some of them are odd, you know, European or Asian breeds.
They're not all AKC breeds, but they are pure breeds; they're not mixed breeds. But there
are other labs that have focused their studies on mixed breeds. So there's one at Cornell
who specializes in village dogs, and he's traveled all over Mexico, South America, Africa,
to the outskirts of town on the garbage dumps, and he and his team sample hundreds and hundreds
and hundreds of mixed breed dogs, and that's, you know, his thing. So, you know, you kind
of can't do everything, and so that's been, you know, where my focus has been, but there
are certainly people doing that. All right, I'm going to let you go. There's lots to see
out in the museum, and I'll be around for questions, and thank you all for your attention.