CARTA: DNA – Neandertal and Denisovan Genomes; Neandertal Genes in Humans; Neandertal Interbreeding

This UCSD TV program is
presented by University

of California Television.

Like what you learn?

Visit our website or follow
us on Facebook and Twitter

to keep up with the
latest programs.

[MUSIC PLAYING]

We are the paradoxical ape.

Bipedal, naked, large-brained,
long the master of fire, tools,

and language, but still trying
to understand ourselves.

Aware that death is inevitable,
yet filled with optimism.

We grow up slowly.

We hand down knowledge.

We empathize and deceive.

We shape the future from
our shared understanding

of the past.

CARTA brings together experts
from diverse disciplines

to exchange insights on who
we are and how we got here,

an exploration made
possible by the generosity

of humans like you.

[MUSIC PLAYING]

Well, thank you very much
for the introduction, Anne,

and to the organizers for
inviting me to participate

in the symposium today,
and also to all of you

for coming back after the break.

So it's nice to
see everyone again.

So I'm really happy to
have the opportunity

to tell you about some
of our recent work

on mapping archaic hominin DNA
in the genomes of modern humans.

And actually, my
talk, you'll see,

is going to be pretty similar
to our first speaker, Shrira.

And in fact, when
he was talking,

I was thinking to myself
how nice it was actually

that some of the
things he was saying

overlapped with what I
was going to talk about,

because we were working
on these projects

completely independently.

We developed very different
statistical methods

to answer the same questions,
and yet, by and large, we

came to many of the
same conclusions.

So I think it engenders
confidence in the things

that we're presenting today.

And my graduate student,
Benjamin Vernot, and I

first became interested in this
question of archaic admixture

a few years ago.

And I think this is actually one
of the most fascinating topics

in all of genetics and genomics
these days, is all of the things

that we've learned from
ancient DNA sequencing.

And one of the more
contentious questions, I think,

in human evolution has
been whether or not

modern humans
mated or hybridized

with archaic humans, like
Neanderthals and Denisovans.

And for many decades, this was
just an acrimonious debate,

and that was largely
because the data didn't

exist to answer the question.

But with technologies developed
by Svante Paabo and some

of the other speakers we've
heard from this morning,

Matias and Kai, in the
not too many years ago,

we were able to get
high-quality genome sequences

from the Neanderthal
and Denisovan genomes.

And this provided
unambiguous evidence

that modern humans and
these archaic humans

did, in fact, hybridize
and exchange genes.

And as Matias talked about
this morning, though,

studying ancient DNA from fossil
still remains really challenging

because you have to find an
appropriate specimen, first

of all, and you have to
hope that the DNA has

been preserved over hundreds
of thousands of years.

So my student and
I thought, well,

if there was gene flow
between modern humans

and archaic humans, maybe we
could excavate ancient DNA,

not directly from fossils, but
indirectly from the genomes

of modern humans.

And to give you a little bit of
an intuition of how this works,

I'd like to argue that a little
bit of archaic introgression

goes a long way.

And so in this schematic,
I'm showing you a picture

of 10 or so individuals.

These aren't random individuals.

These are my colleagues in the
department of genome sciences.

And this is what
happens when you put

your picture on the internet.

So each line here, let's
imagine represents a stretch

of each person's genome.

And from previous work, we
knew that all non-Africans

had about 2% of
their DNA inherited

from Neanderthal ancestors.

And that's what's represented by
these yellow chunks of sequence.

And so what we wanted to
do was develop a method

where we could walk along
an individual's genome

and pull out the parts
that were inherited

from Neanderthal ancestors.

And the key here is that
the 2% of my genome that

was inherited from Neanderthals
might be a little bit

different than your 2%.

So that when we aggregate the
data across many individuals,

we can actually recover
a substantial amount

of the Neanderthal genome.

And actually, what I find most
compelling about this approach

is that as opposed to
sequencing ancient DNA

from a single fossil, by
recovering these all surviving

archaic lineages,
we're potentially

getting data that
was-- or getting

sequences that were inherited
from multiple archaic ancestors.

So we're getting
population level data.

And that will allow
us to make inferences

that are difficult or
impossible to do if you just

have genetic data from
a single individual.

So, I'm not going to talk a lot
about the details of how we scan

along an individual's genome
and look for archaic sequence,

but I did want to give you
a little bit of intuition.

So what are the characteristics
of introgressed archaic sequence

that we look at?

Well, what I'm showing you here
is a simple schematic showing

that Europeans diverged from
Africans about 80,000 years ago

or so.

And what we want to find--

or if we look at lineages
superimposed on this tree,

we can see that what we're
actually interested in finding

are cases like this.

So sequences that are found in
non-Africans that were inherited

from a Neanderthal ancestor.

So what are the
features that we expect

for these types of sequences?

Well, the first is
that in contrast

to two modern human sequences
that have a much more

recent common ancestor, mutation
will have had a long time to act

and accumulate.

We want to find these sequences.

So mutation will have
had a longer time to act

and accumulate on
this lineage compared

to two modern human lineages.

But the other key
feature is that admixture

happened relatively recently.

So in the last 60,000
to 80,000 years or so.

And therefore, the
Neanderthal haplotypes

will still persist over
sizeable genomic regions.

So it's this combination of
highly divergent sequences

that stretch over large
genomic distances that

allow us to accurately
and robustly predict

what are archaic
sequences versus what

are modern human sequences.

And actually, the
approach we're using

is a modification of a
statistic called S star that

was developed by Jeff Wall.

And one of the nice
things about this approach

is it doesn't explicitly use
the Neanderthal or Denisovan

genome when making
the initial inference.

And the really powerful
thing about that is we

can potentially discover
archaic lineages from groups

that we don't even
know about yet.

And actually,
that's a major part

of what we're looking at now.

But that will have to be a
story for a different day.

So what I'm going to
tell you about, though,

is applying this method to
1,500 geographically diverse

individuals.

So whole genome sequences
from about 1,500 people all

throughout the world.

These are largely sequences
from the 1,000 genomes project.

So this is a
publicly-available data set.

But to supplement this, we also
sequenced, in collaboration

with Svante Pablo's group, 35
individuals from Melanesia.

And the idea here was that
we knew from previous work

that these individuals
should have

substantial amounts of both
Neanderthal and Denisovan

sequence.

And if we look at the
amount of archaic ancestry

that we find per
individual, that's

what I'm showing on this slide.

So this shows the
distribution of the amount

of archaic sequence
per individual

in Melanesians, East Asians,
South Asians and Europeans.

And you can see that
Melanesians have, on average,

much more archaic
sequence per individual

compared to some of the
other non-African groups.

And the reason is,
as I just mentioned,

they have substantial amounts of
both Neanderthal and Denisovan

sequence.

And so each row here is an
individual, and the bar plots

correspond to how much
Neanderthal versus Denisovan

sequence each individual has.

And if you look closely, there's
a small amount of sequence

that we label as ambiguous.

This is sequence that we are
confident is archaic in origin,

but we can't distinguish
robustly, at least,

whether or it's
Neanderthal or Denisovan.

So on average, Europeans have
about 50 to 55 mega bases

of archaic sequence
per individual,

and this is largely
Neanderthal in origin.

South Asians have
a little bit more,

East Asians have
a little bit more,

and Melanesians have
about, on average,

a hundred mega basis of archaic
sequence per individual.

So that's 100
million base pairs.

So that's great.

We can identify
archaic sequence.

But I think the really
interesting thing

is the things that we can
potentially learn from it.

So what are the
types of questions

that surviving archaic
lineages allow us to ask?

So, I'm going to tell
you about three things

that we've been interested in.

So the first is
was hybridization

between archaic humans and
modern humans deleterious?

That is, were there
bad consequences?

Conversely, was hybridization
beneficial or were there

some good consequences
of hybridization?

And finally, what
demographic model

is consistent with patterns of
introgressed archaic sequences?

So let's start with
the first question.

Were there deleterious
consequences to hybridization?

And one of the most
striking things

that we found when first
looking at patterns

of Neanderthal sequence
across the genome

is that it's very
heterogeneous distributed.

So I'm showing you sequences
from chromosome 7, 8, and 9.

So the blue ticks
represent places

where we find
Neanderthal sequence

in European individuals, the
red lines indicate places

where we find Neanderthal
sequence in East

Asian individuals, and
the gray lines represent

parts of the genome that
are too repetitive for us

to study and be confident
in the predictions from.

And so if you squint long
enough at this figure,

you can see that
it doesn't appear,

and Shrira mentioned
this earlier,

that patterns of
surviving sequence

are randomly distributed
across the chromosomes.

But you find these
regions that have

been called deserts or
depletions of archaic ancestry

that extend over really
large genomic regions.

And this is
consistent with there

being deleterious consequences
to having Neanderthal sequence

in these regions.

And in fact, when we do
extensive simulations

and try to model
this process, we

see that there's an
excess in the observed

data of these depletions,
or archaic desserts,

compared to simulated data under
neutral models of evolution.

So what does that mean?

It just means basically that
under neutral evolution,

so where there's no
fitness consequences

to the Neanderthal
sequence, we really

wouldn't expect to see desserts
this large in the real data.

So I think this is pretty
compelling evidence

that there was
deleterious fitness

consequences to hybridization.

And what's really
fascinating to me

is that if you also superimpose
Denisovan sequences on top

of this data, you
find that there's

a significant overlap
between Neanderthal desserts

and Denisovan desserts.

So the same places
in the human genome

that are depleted of
Neanderthal sequences

are also depleted of
Denisovan sequences.

And again, this is very
consistent with the idea

that these regions
maybe are harboring

genetic changes that
are very important

to modern human phenotypes.

So for example, the largest
region or the largest depletion

is on chromosome 7.

It's about a 15-megabase desert.

So there's lots of genes.

One of the challenges in
interpreting these regions is

that in a 15-megabase sequence,
there's about 100 genes or so.

So you don't actually know
which one is driving the signal

that you're interested in.

But one thing that caught our
eye, and as Shrira mentioned

this morning, is right in the
middle of this largest desert

is a gene called FOXP2.

And FOXP2 has previously
been associated

with being important
in speech and language.

And in fact, work
from Svante's group

has shown that there's
human-specific mutations

in regulatory regions of FOXP2.

So again, I want
to be careful here.

And we haven't
proven that FOXP2 is

driving this depletion
of Neanderthal sequence

in this region.

But it's really
interesting and I

think these deserts
of archaic ancestry

can help us pinpoint places
in the human genome that

might be important in
modern human evolution.

So the search space
is much narrower now

compared to when we
first did these studies.

Another question that we
were interested in asking

is, well, so it
seems like there were

some deleterious consequences
to hybridization.

Was there also evidence that
maybe some of the sequences we

picked up from
Neanderthals or Denisovans?

Was that beneficial?

And probably the simplest
way to look at this question

is to look at the frequency
of Neanderthal or Denisovan

sequences in modern populations.

And that's what I'm
showing you here.

So each dot
represents a frequency

of either a Neanderthal
or a Denisovan haplotype

in East Asians, Europeans,
Melanesians, and South Asians.

And you can see that
for the most part,

the vast majority of
archaic sequence that

persists in modern human
populations is pretty rare.

So usually less
than 10% frequency.

But there's a number
of regions that

have risen to high frequency.

So 60% in some cases, in some
cases, even a little bit higher.

And we've done extensive
modeling, again,

to try to determine
how likely it

is to see these high
frequency haplotypes

in the absence of selection.

And it turns out
that above 50% or so,

it's actually really
unusual for a haplotype

to randomly drift up to
such high frequencies.

So there's about
100 or so, I think,

really high-confident targets
of adaptive introgression.

And you might wonder, so what
phenotypes were influenced

by adaptive introgression?

And so we knew previously
that a version of a gene

called EPAS1 in certain
Tibetan populations

was inherited from Denisovans.

And it's this gene that allows
them to live at high altitude.

So, there was already some
pretty good a priori evidence

that admixture
with archaic humans

was beneficial for some genes.

And when we look
carefully at these 100

high-frequency
archaic haplotypes,

we see that they are
largely comprised

of genes that can be
categorized into two classes.

One, the immune system.

So many genes that
influence immune

phenotypes, and in
particular, innate immunity.

So the part of our
immune system that deals

with viruses and bacteria.

So that seems to be a very
enriched target or substrate

of adaptive introgression.

And I think you could have
predicted this a priori.

So it's known that the
immune system is often

a target of selection.

But the other category of genes
that actually I would have never

predicted a priori
turns out to be

a number of genes that have
important functions in skin

biology.

So for example, one of these
genes, as Shrira mentioned

is B and C2.

It's a gene called basal
nucleon 2 that has recently

been shown to influence
skin pigmentation

levels in Europeans.

So it's at very high frequency.

Each row here, again,
is an individual

and columns are variant sites.

And these individuals carry
the Neanderthal haplotype.

And you can see that it's a
very high-frequency haplotype

in Europeans, not
found in East Asians.

And finally, and
real quickly, I'm

just going to give you a brief
synopsis on the things we can

learn about demographic models.

And whenever I think
of demographic models,

this image from National
Geographic comes to mind.

I think it's a fascinating
picture, actually.

My kids really like
this too, because they

say I look like him, but
that's a different story.

And so what are the things
that we can try to learn?

Well, we want to know things
like when did hybridization

happen?

How many times did it happen?

Did different populations have
the same or different admixture

histories?

And I have a postdoc,
Joshua Schrieber,

who developed a really
clever method of trying

to infer whether two populations
had the same admixture

history or different
admixture histories.

And so when we apply this
method to pairs of populations

that we analyzed--

the details here
aren't important,

but we can infer this
general picture of--

so this is Europeans, East
Asians, Melanesians, Africans,

Neanderthals, and Denisovans.

And the main point I
want to impress upon you

is that maybe even compared to
as recently as a few years ago,

it seems like the
admixture history

between modern and archaic
humans is much more complex.

And in fact, the data is
consistent with multiple pulses

of admixture between
Neanderthals and modern humans,

and at least one pulse of
admixture with Denisovans.

So, in conclusion,
I've shown you

that substantial amounts of the
Neanderthal and Denisovan genome

remain scattered in the
DNA of modern humans,

that there were fitness
consequences to hybridization,

both good and bad, and
that the history of contact

was much more complex
than previously thought.

And I would like to
thank people in my lab.

So this guy right here in the
middle is Benjamin Vernot.

He was a graduate student who is
now a postdoc with Svante Paabo.

But he, by and large,
did most of the work

that I talked about today.

So with that, I will thank you
and I guess answer questions

after everyone's done.

[APPLAUSE]

[MUSIC PLAYING]

So I would like to start
with acknowledgments.

So the work I'm going
to present is actually

the work of many,
many people who

were involved in
sequencing these two

genomes I'm going to talk about
and help to analyzing them.

I will give some more
credit during the talk.

So let me start by
just introducing

the samples that were used
to generate these sequences.

So both of these
samples were found

in the Denisova cave in the
Altai Mountains in Russia.

And one was actually
a small finger bone

that you see here on top, and
another one was a small toe bone

that you see on the bottom.

And the reason that
we are actually

having two genomes from the same
place has to do with the fact

that the Denisova cave actually
preserves the DNA in these bones

particularly well.

It's one of those
exceptional places where

most of the DNA that we get
out of these very old bones

really comes from
the individual that

died and are not from bacterial
or microbial contamination.

So what this allows
us also to do

is to not just sequence these
genomes to a low coverage,

meaning just a
couple of sequences

from the nuclear genome,
but we can actually

sequence them many times over.

And how this looks
like then is that you

have small sequences
stacking up like this

that are distributed randomly
over the entire genome.

And you always have several
of those for each position.

And so taken together,
you have 30-fold coverage,

meaning at any position in--

on an average position
in the genome,

you will have 30 different
fragments for the finger bone,

and for the toe bone
you have 50 fragments.

So using these genomes
then to actually understand

how they are related,
you see that one

of those two genomes that
we produced from this cave

falls together with
other Neanderthals

that we have sequenced
before to lower coverage.

And we call this the
Altai Neanderthal.

And the second individual--
the nuclear genome

is from a sister
group of Neanderthals

that we call the Denisovan
because they fall outside

of the variation that we
observe of Neanderthals.

So they are more closely
related to Neanderthals,

but they are not looking--
close enoughly related

to deserve to be just
called a Neanderthal,

and we rather call
them Denisovan.

And so one of the
questions that you

might ask yourself is, why
are we actually bothering

with sequencing these
genomes so deeply?

Why do we sequence them
30 or even 50 times over?

And the reason for
this is the fact

that we are diploid organism.

So we are actually having
each chromosome twice.

One complete set
are inherited from--

is inherited from the mother,
and another complete set

of chromosomes is
inherited from the father.

And so one of the
things that you

can do when you have
so many fragments

and so many sequences is you can
call confidently the differences

between these two
copies that you have

from the mother and the father.

And this is really the reason
why we are sequencing it

so deeply so that we
can call the differences

between these chromosomes.

And one of the most easy
analysis that you can actually

do once you have
sequenced so deeply

and call these differences
between the chromosomes

is to just ask how different
are they on average?

So this is called
heterozygosity.

And you can actually put
this into perspective

by also showing,
as in this plot,

the level of
heterozygosity, so the level

of differences between the
chromosomes, in modern humans,

in present-day modern humans.

And we have some
individuals from Africa here

and some individuals
from outside of Africa.

And what you can see is that
Africans have about 1 in 1,000

differences between the
chromosomes that they inherited

from the mother and the father,
while non-Africans have between

6 and 8 in 10,000.

And the archaics are much
reduced compared to both

of these present-day human
populations or present-day human

regions, and they are at a
level of 2 to 3 in 10,000.

And there's even a quite
significant difference

between the two archaics
in that the Denisova

is higher than the Neanderthal.

So the Neanderthal
is further reduced.

So one can actually look
into this in more detail

by looking over the chromosomes.

So just going in a small
window over the chromosomes

and just counting the
differences that you observe.

And we have done this here for a
French individual, the Denisova

and the Altai Neanderthal.

And what you can see is that
the level of heterozygosity,

so the differences
between the chromosomes,

varies over the genome.

But there's one thing that
is very special, and that

is that the Altai Neanderthal
has this very long stretch here,

for instance, on chromosome 21--

there are other stretches like
this on the other chromosomes--

where there's hardly any
difference between the two

parental copies.

And so now what one can
do is one can actually

take the size of these stretches
and how much of the genome

is actually residing
in those stretches

to calculate back
how closely related

the parents would have to be to
generate stretches like this.

And this is an analysis that
Flora Fay in Monty Slatkin's lab

in Berkeley was carrying out for
for the analysis of the Altai

Neanderthal genome.

And what she found
is that there are

several different relationships
between the parents,

possible that would
actually generate exactly

the patterns that we see.

And so I guess one
easy way to say

this is that the parents
of this individual

would have to be
at least related

on the level of half siblings
to generate these patterns.

So they were closely related.

And then you can take
it a step further

and you just take
your prediction

of how much you
would actually expect

in terms of long
stretches that are looking

like this, almost identical.

And you just ask, if
I would subtract now,

based on what I
understood, the family

relationship of the
parents would be,

if I subtract this away, is
there actually anything left?

And this is, in fact, the case.

So for the Altai
Neanderthal, you still

see an excess over the stretches
that you see in the Denisova

and in modern humans.

And this actually means that
this is not just a singular

event that just once
happened, that just by chance,

the parents were
closely related,

but also further
back in the past,

they were probably
closely related ancestors.

So another topic that I
would like to talk about

is archaic admixture.

And so we already heard
about archaic admixture

from Neanderthals and
Denisovans into modern humans.

What I would like to talk about
is really archaic admixture

between both archaics.

But before I get to
this, I would actually

like to go a little bit
deeper into how we actually

know what signal we have to look
for to understand that there

is really admixture.

And so as a very simple
way of depicting this,

just imagine that you
have a certain individual.

And of course, as I
already explained,

every chromosome has two copies.

So this individual has these two
copies of a certain chromosome,

an arbitrary one.

Of course, you can go
back to his parents,

and one of those copies
will come from the father,

and the other one will
come from the mother.

So I can paint them
now blue and red.

But I can also go a step
further and actually

paint them according to whether
they come from the grandparents

or from which
grandparent they come.

And what you see
in this picture now

is that there is actually a
process called recombination

that is actually mixing up
these different chromosomes

in the parents of
the individual.

So now you have these
random stretches

from all the grandparents that
are painting these chromosomes.

And of course, you can take
another step, another step,

another step.

And so essentially
what this means

is that you're breaking
up the ancestry-- when

you go through
the ancestors, you

jump between
different ancestries

when you go over the chromosome.

So you change which
ancestor's genome you look at.

And so when you repeat this
process for a very long time,

and let's say you have ancestry
from one population for most

of the genome, but you
have a couple of ancestors

from a different human group
hiding among your ancestors,

then what the most
common outcome will be

is that you will have these
short stretches, where

one of the chromosomes
actually shows

this ancestry of this
other human group,

while the other
chromosome is actually

looking like the chromosome
of the majority of the groups

because these stretches
will be randomly placed

on your two chromosomes.

And now we can actually
use the Altai Neanderthal

and the Denisova to
find out whether there

is any Neanderthal
ancestry in the Denisova,

or whether there is
any Denisovan ancestry

in the Neanderthal
individual that we sequenced.

And so in one direction,
just showing--

so if Neanderthals would
contribute to the Denisovan,

what we would expect is that
there are some stretches where

the Denisovan looks very
much like a Neanderthal,

but on the other chromosome, we
would expect that it actually

looks like a normal Denisovan.

That means that the two
chromosomes are actually

very different.

And so the prediction
that this makes

is that if you go
to regions where--

shown here on the
left-hand side,

so the x scale is giving
you how closely related

you can this-- make any
particular window that you look

through, how closely related
you can actually make that

to the other archaic.

So when you have windows
where the Altai Neanderthal is

very closely related to the
Denisovan, shown here in blue,

you actually see no effect.

But if you look in the
same for the Denisovans,

when the Denisova can be made
very closely related-- or looks

actually very closely related
to the Altai Neanderthal,

you see that the two
chromosomes are very different,

and this is shown here
in red at this position.

So this is a hallmark sign
that there is actually,

among the ancestors
of the Denisovans,

some Neanderthal ancestry.

The last signal I
want to talk about

is actually the one of
unknown archaic material

that we found in the Denisovan.

And so the first signal
that we saw for that

is really just when you look
for divergence to Africa,

so that's nothing else than just
looking for how many differences

we observe, we actually see
when we take larger windows

and we just compare
to an African,

that the Denisova is always
a little bit more different

than the Altai Neanderthal.

So these two
distributions that you

see here, they are-- the
one for the Denisova in blue

is slightly shifted
to the right.

And you can look
even deeper into this

by actually looking at
different allele frequencies

and divide up your comparison by
how many Africans actually carry

a certain derived variant,
meaning a new variant that

occurred some time after
the split from chimpanzee.

And then all Africans
are the same,

you actually see that the
signal is the strongest.

So you see the most differences.

And in an analysis that also
Monty Slatkin's lab carried out

in Berkeley with Fernando
Racimo, what they did

was essentially taking the
signals that I just described,

and they tried three
different models

to actually compare how
this could come about.

And so the first model
assumes that there

was gene flow from Neanderthals
into the common ancestor

of all modern humans.

The second model is assuming
that all modern humans actually

gave some material
to the Neanderthals.

And so these first two
models are essentially

trying to explain
how you could make

the Neanderthals and the modern
humans more closely related.

And the last model is one where
you have some lineage that we

haven't observed, so we
don't know what it is, that

contributed to the Denisovans.

And that would make the
Denisovans more distantly

related to modern humans.

And so in most
comparisons, the model 3

was actually the
best explanation

that we could find for the data.

And so we believe that there
is this super archaic admixture

of some very deeply divergent
lineage into the Denisovan.

So what I would like
to say in the end is--

or what I would like
to show in the end

is really just a general
overview of the different gene

flows that we have now observed.

And this picture is
not quite complete yet.

So what you can
see is that we have

the ancestry of these deeply
divergent ancestor that

contributed to Denisovans.

We have the
Neanderthal admixture

into the modern humans, we have
contributions from Denisovans

to modern humans, and
the Neanderthal admixture

into the Denisovans that
I just talked about.

And they are-- it seems that
there is-- by now there are also

other publications that say
that there are contributions

to Africans and so on and so on.

And so I think what this all
means, when you sum it up,

is that these different
types of admixtures

are actually something
that is quite common.

So it actually happens
quite a lot in the past,

and that is something
that is really

a transition in our thinking,
because originally, I

think we were all very skeptical
that there was actually

any admixture between
these archaic groups.

And with this, I
would like to end

and say thank you
for your attention.

[APPLAUSE]

[MUSIC PLAYING]

Thank you very much.

So I couldn't have
asked for a better

introduction for what I'm going
to talk to you about here today.

We've heard from several
of the previous speakers

about the genetic legacy of
interbreeding with Neanderthals.

But I'm very interested in
understanding what, if anything,

is the phenotypic legacy in
modern human populations?

Is this Neanderthal DNA that
remains in us, is it functional?

And if so, what
function does it have?

And so as we've seen, thanks
to the pioneering work

of many of these
previous speakers,

we know that
Neanderthal DNA remains

in certain modern
human populations.

And if we look at a schematic
of a human chromosome here,

you can think of this as a long
string of As, Ts Cs and Gs.

I've colored in blue all the
locations where we've ever

observed someone living
today to have Neanderthal DNA

in their genome.

And if you look across
many, many thousands

of European and
Asian individuals,

you'll see that on average,
around 2% of their genomes

are derived from
Neanderthal interbreeding.

As we've heard, different
people will have a different 2%.

My 2% is different than Ed's
2%, is different than Anne's 2%.

And I want you to
remember that because this

is a really important feature
that we're going to use later

to try to understand the
function of these different bits

of Neanderthal DNA that
remain in our genomes.

And at some parts
of our genome are

more likely to retain
Neanderthal DNA than others.

So in one extreme, we see
these Neanderthal deserts,

like the position here
on the right=hand side,

where we've never observed
anyone to have Neanderthal DNA.

And then on the
left-hand side, we

have the other extreme
where we have up to 60%

of European individuals.

If you went out and sequenced,
a bunch of European people

would have Neanderthal
DNA at that location.

And so ultimately, this
suggests that Neanderthal DNA

had an influence
on our ancestors

after the interbreeding.

In some cases, perhaps
positive, in other cases,

perhaps negative.

And so for me, this
raised a very big question

that I really wanted to
answer is, OK, so then

what is the phenotypic legacy of
this Neanderthal interbreeding

and the DNA that remains
from it in modern humans?

And so I hope if you remember
nothing else from my talk,

really just two main points.

The first is that indeed,
interbreeding with Neanderthals

has left a phenotypic
legacy in modern humans.

And the way I'm going
to go about trying

to show what that
legacy has been

is using a new type of resource
that's just becoming available,

and that's of large
clinical biobanks

with electronic medical
records from patients,

from hospitals linked
to genetic information.

And this is a really,
really powerful resource

for studying the
genetics of disease,

but I also think it's a really,
really powerful resource

for studying the genetics
of our recent evolution.

And so if you want
to, you can go

to sleep now and just
remember those two things

and I won't blame you.

So basically, we got the
idea for this project

because I collaborate with a
big national consortium called

the Electronic Medical
Records and Genomics Network.

And what this is
is a collaboration

of about 10 academic hospitals
from across the nation

that have electronic medical
record systems implemented

in their hospitals, and
also genetic information

from those patients linked
to their electronic medical

records.

And so this looks
a little something

like this, where on
the left-hand side

we have John Doe's
patient record.

He's been coming to the
hospital and seeing doctors,

let's say, for
the last 10 years.

And we've got records of
all those events and all

the treatments he's received
in that electronic form.

And then some day John comes in
to have blood drawn and he says,

"Yeah, actually, it'd be OK if
you use any leftover material

from this blood draw for
basic medical research."

And if he's
consented to do that,

then all that
information is sent

through a de-identifying
process where

all the identifying
information is removed

from that electronic
medical record,

but the basics of the treatment
history are maintained.

And then the blood
sample is also

passed through and biobanked
and given an ID that links it up

to that anonymized version of
the electronic medical record.

And now this is really
powerful because it

enables us to do
genetic association

testing on a very large scale.

So what is genetic
association testing?

Well, we can--
let's imagine we've

got a number of patients
here for which we have

these biobanked blood samples.

And let's say we're interested
in studying something

about their genetics.

Well, we can look at
these blood samples

and see at one given
position in their genome

whether or not they
have an A, T, C, or G.

And so in this example, patient
1 has an A, patient 2 has an A,

and then patient N has a G.

And let's say we're also
interested in heart disease

and whether or not this
particular location

in those patients genome has any
effect on their risk for heart

disease.

What we can do is then go look
in their electronic medical

record and say, all right,
well, has this person ever

been treated for heart disease?

And let's say in this case we
find that, yes, patients 1 and 2

have, and then
patient N has not.

And once we have
that information,

we can perform statistical
tests for association

between these individual's
DNA at that given

position in their genome,
and whether or not they've

ever been treated
for heart disease.

And so in this
simplistic example,

we might say that
yes, having an A

at this location in your
genome increases your risk

for heart disease.

Now, of course,
we don't normally

do this on three people, we
do this on tens of thousands

of people to try to find
significant associations

between regions of our
genome and disease.

Now, this is all well and
good, but let's say we're

interested in another disease.

Let's say we're interested in
arthritis and the genetic basis

for arthritis.

Well, if we didn't have this
electronic medical record

system, we'd have to go out and
collect a whole other cohort

of people that had arthritis
and then some matched control

people that didn't have
arthritis and then genotype

them and then test whether
or not those genetic loci had

any effect on the risk.

But because we have the
electronic medical record

system, we can instead
just go look in the record

and say, all right, let's find
a new set of cases and controls

for arthritis and perform
genetic association tests,

again, on the genetic
information we already have.

So that's all well and
good, but we're here

because we care about human
origins and human evolution.

So let's get back to that.

How can we use this kind of
data to answer this question

about the effects of the
Neanderthal DNA that remains

in modern human populations?

And so what we did
was to start with data

from this large eMERGE,
Electronic Medical Records

and Genomics Network,
from across the country.

We got data for about
28,000 patients from across

the country.

And we first looked
at their genotypes.

We first found genetic
information from about 600,000

positions across their genomes.

And so you can think of
this as a string, again,

of about 600,000 As, Ts, Cs, and
Gs that we've associated with

each one of these patients.

And then what we realized we
could do was use these great

high-quality maps of
Neanderthal DNA that remain in--

remains in modern
human populations

that you've heard about
from Shrira and Josh.

And so we could
look at those maps

and then intersect them
with our own patients

and apply those techniques
to our patients genomes

and identify regions where each
patient had Neanderthal DNA.

And so we could do this for
about 1,500 of these positions

in these patient's genomes.

And we can see where some
may have Neanderthal DNA

and others may not.

And then finally, the last
piece, as I indicated before,

comes from using these
electronic medical record

data to define a set
of phenotypes or traits

for each of these patients.

We can ask for hundreds
of different phenotypes,

covering the whole spectrum
of things you might

be treated for by a doctor.

Whether or not each of these
people either had that disease,

they were a case, or
they were a control,

or we couldn't
really figure it out

and we should leave them
out of the analysis.

And so then using this matrix of
data of genetic data annotated

with Neanderthal
ancestry and then

many, many different
phenotypes, we

were able to start testing for
the effects of Neanderthal DNA

on a much broader scale than
really had been possible before.

And so before I get into
what we actually find,

I'll try to be a good
scientist and think

about what we would expect to
find before actually running

the experiment.

And so what did we expect?

Now, as modern humans migrated
out of Africa where they first

appeared, they
encountered a number

of different environments.

So they encountered different
climates, different levels

of sun exposure,
different temperatures,

different seasonal patterns.

They also encountered
different animals and plants

that led to different diets,
and very importantly, they also

encountered different pathogens.

And so it's been
proposed that perhaps

by interbreeding with
Neanderthals and Denisovans,

and perhaps other
archaic human forms

that had been living in these
environments for hundreds

of thousands of
years, in many cases,

before anatomically
modern human groups

ever arrived there,
perhaps there really

was some adaptive benefit
you could get from spending

a night with a Neanderthal.

Maybe that was not
such a bad trade off.

But this is really a hypothesis.

This hasn't been shown at all.

So under this
hypothesis, we might

expect that the Neanderthal DNA
that could have been adaptive

in our modern human populations
would have been influencing

human traits that are
involved in interactions

with the environment.

So things like our
immune system, of course,

be one of the most important.

But our skin perhaps,
and perhaps also

our metabolism or other
traits related to our diet.

And so we also expected that
we might see some effects

on our bone or
skeletal structure

because we also know about
many important differences

or many very easily
detectable differences

between the bones of
anatomically modern humans

and Neanderthals.

So those are some of the
things we were expecting

as we went into this analysis.

So what did we find?

And now in doing this analysis,
we decided to split up our data,

our 28,000 individuals,
into two different sets,

a discovery cohort of
about 13,500 individuals,

which we'd run an
initial analysis,

and then a replication cohort in
which we would try to replicate

anything that we found
in that first cohort.

So in the discovery, I'm
going to show you just some

of the top associations we
found between Neanderthal DNA

and potential phenotypes
in European ancestry,

anatomically-modern
human populations.

And so when I saw this, I
almost couldn't believe it

because what do
we see at the top?

We see osteoporosis,
a bone trait.

Then we see
hypercoagulable state.

So what is that?

That's just blood clotting.

Your blood's too thick.

It clots too much, which can
lead to all sorts of problems.

Then we see protein calorie
malnutrition, a metabolic trait.

And so this is really
surprisingly matching

what we expected.

But before I go too
far into interpreting

these, let's talk about
that replication analysis

I mentioned.

So what we did here is we looked
at the other 14,500 individuals

we left out of the initial
analysis and tested to see

whether we saw consistent
effects in that group.

And so luckily for four
of these top associations

I'm telling about, we did
see something consistent.

We did see a consistent effect.

Unfortunately, the osteoporosis
one did not replicate there.

And I should say,
just as an aside,

I don't think that necessarily
means it's not true,

but it's notoriously difficult
sometimes to replicate these

genetic associations and
we're following that up

in some other cohorts.

So let's focus on these
four that did replicate.

So first we have this
hypercoagulable state

association that I already
talked a little bit about.

So this means that your blood
coagulates very quickly.

And this is actually
a very important part

of the early immune response.

The coagulation
factors are really

some of the first proteins that
pathogens interact with when

they come into your body.

And so this really
fits in with this idea

of the potential
immune benefits.

And we've looked into
the molecular basis

for this association
and we've actually

been able to show that the
Neanderthal DNA nearby--

sorry, this Neanderthal DNA that
is associated with increased

coagulation increases the level
of several nearby coagulation

factors in your blood.

So we have a very compelling
molecular mechanism

for how that might be happening.

And now, I'm sure by now you've
read the rest of this list

and seen one that's a little
bit more difficult to interpret.

And that's tobacco use disorder.

And so that really just
means addiction to nicotine.

And so I think should we
be thinking about this?

Were Neanderthals sitting
around outside of caves smoking?

And I want to say
unequivocally, no.

No, we cannot say this.

You should not say this,
you should not think this.

This extreme example highlights
a really important point,

that the effects of genetic
variation in modern environments

may not actually reflect its
effects 50,000 years ago against

a very different genetic
background and Neanderthals,

or in early human
Neanderthal hybrids.

And on top of that, of course,
tobacco is a new-world plant.

They didn't really have nicotine
existing in their environment.

But what this does tell
us is that Neanderthal DNA

in modern humans is influencing
a system in our body

that is now, in
modern environments,

relevant to this trait.

And in particular, this bit of
Neanderthal DNA is very nearby,

a transporter for
a neurotransmitter

called GABA that's involved in
all sorts of important processes

in the brain and even may have
a role in circadian processes.

So we don't really
know what might have

been behind this association.

Now, just to move on, I want
to tell you about one more

analysis that we did.

So in that first
set of tests, we

were testing for the effect
of one bit of Neanderthal DNA

with one trait in
a human population.

That we wondered,
well, what if we

looked at all the
Neanderthal DNA

that a person might or
might not have in aggregate,

and ask whether or not
that could predict,

better predict someone's
risk for a disease.

And so we did an
analysis of that.

And again, we found several
very interesting associations

that replicated.

And now I think this top one
is really, really fascinating.

It's Neanderthal DNA-- if I know
your Neanderthal DNA complement,

I can better predict your
risk for actinic keratosis.

And this is-- in case you don't
know, this is a skin disease.

It's not terribly serious.

It's often seen in
fair-skinned people

after long-term sun exposure.

And it's caused
by malfunctioning

of a gene-- of a type of cell in
your skin called keratinocytes.

And I find this so fascinating
for really several reasons

because keratinocytes, one
of their main functions

is protecting our skin
from UV radiation.

So again, a very important
environmental difference

between Africa and other
non-African environments.

But they're also really
intimately involved

in early stages of the
innate immune response

and signaling for the activation
of other immune factors.

When we look at patterns of
where Neanderthal DNA falls

in our genome, we
see that many of

the Neanderthal-- high-frequency
Neanderthal bits of DNA

are nearby genes that are
involved in keratin biology.

And so this is taking
it to the next step

and showing, not only
is it enriched nearby

those genes, but actually
in modern populations,

it's having an effect
on a phenotype that's

very relevant to keratin.

But again here we'll see there's
a second confusing, or at least

more complicated to
interpret association

that we need to think about,
and that's depression.

And so again, I really
want to be very clear

that this is not what we
should be thinking about.

Neanderthals, we cannot
say they were depressed.

We cannot blame them for
any depression we have.

These are very
complex phenotypes

with major
environmental components

and many other
genetic components.

And the Neanderthal
influence is really

quite modest in the whole
constellation of all

the contributions to them.

So in conclusion, I
want you to remember

that interbreeding
with Neanderthals

has indeed left a phenotypic
legacy in modern humans.

And in particular,
it's left effects

on many different
systems in our bodies.

Our immune systems, our skin,
our metabolism, and in fact,

even likely, our brains.

And so, I think largely because
of the nature of the data sets

we've been looking at,
we've found many cases

where the Neanderthal DNA has
a mildly deleterious effect

in modern environments.

But again, I want to remind
you that's not necessarily true

50,000 years ago when this
interbreeding likely occurred.

And so one of the main
challenges going forward

is trying to understand
what knowing something

about Neanderthal DNA
in a modern environment

can actually tell us about
what was happening back then.

And so then the second point
I wanted you to remember

is that this was all enabled by
using a new type of resource.

These large-scale databases
of tens or hundreds

of thousands of
electronic medical records

from patients linked up
to genetic information.

And so, I think just as the
ability to sequence people's

DNA at large scale
has dramatically

changed our understanding of the
genetic basis of human evolution

over the past 5 or
10 years, thanks

to many of the speakers
in the symposium,

I think that leveraging
these sorts of data

and these sorts of
projects that are popping

up all over the
world will allow us

to do the same thing
for the phenotypic basis

of recent human evolution.

And so with that, I would
like to say thank you

all very much for listening and
thank all of my collaborators.

And yeah, thanks.

[APPLAUSE]

[MUSIC PLAYING]