- [Anna] Welcome again. Good afternoon. This is our newest installment of the Units of Life Seminar Series. As a reminder, given the fact that many
of you may be new to this and it's a relatively new seminar series, we started this with the idea that maybe we could have a
place where we can convene to discuss deeper mechanistic biology. Here at the Broad, we're really great at using
a lot of omic technologies, and really doing amazing work with that. But sometimes it's really interesting also to see how we can go deeper into some of these biological
mechanisms and learn more. And so, we thought that maybe
crosscutting across all areas, we could have a seminar series that's more dedicated to that. So, as with many things around here, this is also an experiment
which I sort of took on, and Gillian Shaw has been my
partner in organizing this. And so, this is really the second year that we're running this seminar series, approximately once a month, sometimes every couple of months. And as you can see from the slides that are flashing through, we have some really amazing speakers, including Paul Cohen today and some incredible speakers coming up. John Brugie, Helen Hobbs,
(indistinct) is coming for a special seminar in
combination with MIT and Harvard, and it's going to be on a
Tuesday for that particular time. And then Anna Maria
Corvo and Arlene Sharpe to close out the year, and really amazing speakers
lined up for next year as well, I promise you. And so, without further ado, I wanted to actually
introduce today's speaker. So, Paul Cohen is the Albert
Resnick Associate Professor at the Rockefeller University. He is actually the Head of the Laboratory of
Molecular Metabolism there, and he's, however, no stranger to Boston. As I learned, actually, he
grew up not so far from here, but he also received his
AB from Harvard College and then went to New York to attend the tri-institutional
MB PhD program, which includes the Rockefeller. He did his PhD with Jeff Friedman there and then went onto Columbia
for his residency training, and then came back to Boston,
to the Brigham and the Farber, for fellowship training
and post-doctoral work, and then back to New York to join the faculty at the Rockefeller where, as I mentioned, he now heads the Laboratory
for Molecular Metabolism. Paul has actually done
some really interesting and remarkable work in adipocyte biology and really understanding adipocytes from the perspective of both
his cardiology training, but also links to cancer, which makes him a very, you know, it's a very interesting
and unusual combination of trying to understand, you know, the intercellular crosstalk
and how this works with different comorbidities
related to obesity. And so, I think that we're in for a treat. So, Paul, thank you so much for joining us and we look forward to your seminar. Oh, I should mention before I go that there is a Q and A for those online, and Gillian has kindly
agreed to moderate that. So we will be monitoring
that for questions. And of course, everyone in the room, please feel free to ask questions. Because we're doing this in a hybrid form, we have mics. Otherwise, our friends online
would not be able to hear us. So try to make an effort to go to the mic and we'll go from there. Thank you so much. - [Paul] It's been a great
day visiting with old friends and some colleagues I hadn't met before but whose work I've followed
and admired for a long time. And, as Anna mentioned, I was in this area for nearly eight years and actually lived just a few blocks away, so it's kind of neat to walk
around my old neighborhood. At any rate, please feel free to interrupt
me if you have any questions. There should also be plenty of time for discussion at the end, and I'll try to point to things
on the screen here, as well. So for me, the backdrop to our work is this huge
biomedical problem of obesity. And as you know, this is a problem that the United States
is particularly good at. And by 2030 it's estimated
that 50 percent of adults in the United States is going to be obese. But this is actually no
longer just a problem of industrialized nations. It's a global problem, as well. And for the first time
in our history as humans, diseases due to overnutrition
are actually more prevalent than diseases due to undernutrition. And so, of course, obesity per se, or the weight on the
scale, is not the issue. It's all of these comorbid
diseases that come with obesity: heart disease, diabetes, kidney disease, many different kinds of cancer, and over the past few years as
we've unfortunately learned, worse outcomes with COVID-19. And just to kind of highlight
at a more personal level, what got me interested in this question, I now see patients at Sloan Kettering, which is a cancer center,
with a focus on patients with obesity and cardiometabolic
disease in cancer and it's surprising
how often we see people with phenotypes like through here. The first is a 75-year old man
with obesity, hypertension, hyperlipidemia, coronary artery disease, atrial fibrillation, and
congestive heart failure with rectal cancer. And the second is a 52-year old woman with an array of comorbidities, including obesity, coronary disease, congestive heart failure. She's had lymphoma, breast cancer, and newly-diagnosed lung cancer. And I don't mean to imply that their health problems
are solely due to their BMI, but clearly it's a major
contributor to their pathology. And so, as we think about
these comorbid conditions, the fundamental question of my group has been to
understand how this occurs. And we feel that the answers can be found, at least in part, by studying the biology
of fat or adipose tissue. And so, our work focuses in two areas. One is the biology of white fat. These are electron micrographs of white or brown adipose tissue and we'll discuss each of
these a little bit more. And so, we are interested in
what mechanisms link obesity to these diseases and in particular, how does white fat contribute
to obesity associated disease. And then secondly, and this will be my focus of today's talk, might brown fat actually be an answer, or a way to unlink obesity
from these diseases? And so we've been asking, can it protect from
obesity associated diseases and if so, how would that work? And so, our approaches really
are more focused on a question than a particular methodology. And so as a result, they span from cell biology
to tissue homeostasis into organ crosstalk and
translation into humans, and I'll touch on each of
these in my talk today. So just to highlight a little bit more about the cell biology of fat cells, white fat cells have evolved
to store access calories in the form of lipid. They have these large,
unilocular lipid droplets which store energy as triglyceride. Very few mitochondria and when
energy sources are scarce, the lipid can be mobilized by lipolysis to provide free fatty
acids as a source of fuel. Brown fat cells, on the other hand, have these small,
multilocular lipid droplets. They have a very high
number of mitochondria. In fact, it's the iron in the mitochondria that literally gives tissues, containing brown adipocytes a brown color. And rather than storing energy, these cells can actually
dissipate energy as heat via a process called
uncoupled thermogenesis. So this just illustrates
how that process works. You probably remember from
your biochemistry class days that in oxidative
metabolism, what happens is that reducing equivalents
enter the mitochondria and are passed along the
electron transport chain. A consequence of this is the generation of the proton gradient across
the mitochondrial membrane. And normally, that proton gradient is used to power the synthesis of
ATP through complex five or ATP synthase. And what is unique about
brown adipocytes is that they express a protein called UCP one or uncoupling protein one, which sits in the inner
mitochondrial membrane and is a proton fatty acid symporter. And so what this protein
does is it allows the protons to leak back across the inner membrane, thereby dissipating the gradient and creating a futile cycle. Because the gradient is lost, all of the upstream steps are up-regulated to try and reestablish this gradient, which of course can never happen. And because many of those
reactions are exothermic, the end byproduct of that
is the generation of heat. And we know that these cells
not only can generate heat, but from mouse studies
dating back many years ago from Brad Lowell and others, if you ablate all UCP one
positive cells in an animal, this results in a phenotype of
cold intolerance and obesity. And it makes sense that these cells have a role in metabolism because in order to do
this, they need fuel and they help clear a number of toxic metabolites from the blood, including glucose, free fatty acids, and branch and amino acids. So in terms of its evolutionary role, we think that we developed brown fat to protect against hypothermia. And for mammals, brown fat
is a mammalian adaptation. Hypothermia is a major
threat to early life because we're born small and hairless with a fairly large surface
area relative to our mass. And so here, I'm just showing you, oh, that's not on my screen so I don't have to do anything. Here, I'm just showing you
a thermal camera imaging of a litter of newborn mouse pups. And what you can see, these are some newborns
all lying in a pile on top of one another. This pup has gone venturing
away from its brethren and you can see this very
strong heat-emitting signal on its back between its shoulders and this is the so-called
developmentally preformed, interscapular brown fat. And this same depot of
brown fat is present in newborn humans in an
interscapular location. We also know, though, that
mammals have inducible brown fat. So in response to cold,
fat has the ability to turn on a program of thermogenesis in a type of cell called beige fat cells. And what I'm showing you
here are dissections of mice where all you see is the fat
and the associated organs. And this is from a mouse
housed at 28 degrees Celsius. You can see the developmentally
preformed brown fat, but after housing the animal
at six degrees Celsius where it can happily live for
an indefinite period of time, you can see the clear change
in the appearance of the fat, and this represents the induction of these multilocular beige adipocytes with a high number of mitochondria, giving the tissues the brown color. And so, I mentioned there, the second type of inducible brown adipocyte
called the beige adipocyte. These cells have many
of the same properties as brown adipocytes, with one of the main distinctions being that their activity is highly inducible. So in response to cold,
beta-adrenergic agonists, exercise or thiazolidinedione drugs, these cells can increase
their thermogenic activity. And, importantly, we
know that the brown fat that we can detect in humans shares a
transcriptional signature that's very similar to rodent beige fat, and so we think these are a
particularly useful model system for studying thermogenic fat. At a transcriptional level, the identity of these cells is dictated by a transcriptional co-regulatory
protein called PRDM16. And in early work in
Bruce Spiegelman's lab or out of the post-doc
done by Patrick Seal and Shingo Kajimura, they
showed that PRRDM16 is required for the formation of
thermogenic fat cells. So if you ectopically
express PRDM16 in a fat cell, it gives it the molecular, morphologic and bioenergetic properties
of the thermogenic fat cell. And the way that PRRDM16
works in fat cells is not by binding DNA itself, but by contacting other
transcriptional components, including PPAR gamma, PPC one
L, and the mediator complex to activate the expression of brown and beige fat selective genes. And then it can also bind to CPDP and histo-modifying enzymes to modify the chromatin to
suppress white fat genes. So it works in two ways
to dictate the development and maintenance of these cells. And so, in work that I did as a postdoc, I generated and
characterized a mouse model that has an adipocyte
specific knockout of PRDM16. And because this knockout
is done postnatally, the mice specifically lose
their inducible beige fat, but not their preformed ground fat. And so as a result, they do not have any
difficulty tolerating the cold but they do have many features
akin to metabolic syndrome and humans including more
visceral fat, mild obesity, insulin resistance, and hepatic steatosis. And so, this finding and many of the others that I mentioned, as well as a whole body of
work that I don't have time to discuss today, has led to this great
interest in the field as to whether we could perhaps
target thermogenic fat for therapeutic benefit, either as an anti-obesity
agent or as a way to mitigate against obesity
associated diseases. However, until about a decade ago, I think most people would've told you that was a waste of time. And the reason for that
was because the feeling was that thermogenic fat was really important in small mammals like mice
and in newborn humans, but that in adults, brown fat atrophied and was no longer relevant. And that notion really was
overturned with a series of papers published in the New
England Journal of Medicine in 2009 showing that adult humans do in fact have cold inducible brown fat. So what I'm showing you
here is an FDG PET CT scan. These scans measure the
uptake of radiolabeled glucose and then the CT scan can allow you to see the imaging
characteristics of the tissues that take up the glucose. And in this particular study, young healthy subjects were imaged either at thermo-neutrality, which
is how most of us live, or after just a couple of
hours of moderate cold exposure and you can now see the appearance
of these FDG avid tissues along the neck and cervical spine. And in one of these tissues, subjects actually allowed
themselves to have biopsies taken of these tissues, which is
bearing because these are close to some vital vessels. But the tissue of pain showed that these cells here have the morphologic and molecular
characteristics of brown fat. And so, most of our work since
then has come in two forms, the field's work I should say. So either retrospective studies, looking at series of PET
scans and linking them to clinical data and usually numbering up
to a few hundred patients, or small prospective
studies in which roughly 10 to 20 subjects, typically
healthy young male subjects, are prospectively exposed to cold and then some metabolic
readout is studied. And so, these studies have
been incredibly useful in suggesting that brown fat
promotes insulin sensitization and glucose lowering but
because of their size, they weren't really powered
to look more comprehensively at the health effects
associated with brown fat. And so, that question
motivated Tobias Becher, a former fellow in my
lab, who was interested in this broad question
of does brown fat protect against obesity related comorbidities? And to do that, he took
advantage of clinical PET scans. So unlike the research scans I showed you, where subjects are deliberately
exposed to cold to try and bring out brown fat, in
clinical scans, if anything, patients are kept warm to
suppress brown fat activity because these scans are done to diagnose or track the progression of cancer and you don't want any extra signal. And what is clear if you start
looking at these scans is that there's a tremendous
natural variation in brown fat. So you see this individual
has quite a lot of brown fat, whereas this individual
seems to have hardly any or perhaps none at all. And so, Tobias and I quickly realized that because I had a clinical
role at Sloan Kettering where they do somewhere between 10 and 20,000 PET scans a year, this would be a rich resource
for looking at associations between brown fat and
health outcomes in humans. And so, what he did was he
retrospectively reviewed about a decade worth of studies and because it's their habit
there to always comment in their reports as to whether
or not brown fat is seen or not, we could look
through all of the reports and segregate people into
either brown fat positive or brown fat negative, and then we could associate
their brown fat status with all of the other information and their electronic health records. And so in total, this gave us a cohort of
more than 50,000 patients with over 130,000 scans. The discrepancy is because a number of patients have multiple scans. So this was at least an order
of magnitude greater in size than any series we were aware of. And so, the other thing
that we did then was to try and develop a matched cohort. So of course, an important consideration
when one is doing these kinds of retrospective studies is you need to consider different covariates. And it was already well established
that brown fat associate with age, sex, BMI and
outdoor temperature, meaning brown fat is less
prevalent as we get older, unfortunately, less prevalent in males, less prevalent with higher
BMI and more prevalent in colder outdoor temperature. And so, what we did was we
took people with brown fat and matched them roughly one to two based on these four covariates to give us a matched cohort
of about 5,000 people with brown fat and about 10,000
people without brown fat. And when we then linked
that to the diagnosis codes and their electronic health records, we saw that the odds of a variety of cardiometabolic diseases
were significantly reduced in people with brown fat. So most strikingly type two diabetes, which had been described before, but also a number of associations that had not been described before, such as coronary artery disease, congestive heart failure,
and hypertension. And importantly, because
this is a matched cohort, these effects are obesity independent. So this has nothing to do
with body weight or BMI. This is based on brown fat status. And if we look now by
segregating people by BMI, and these are the different diagnoses but I'll just draw your attention
to type two diabetes here, if we group people into
categories based on BMI, either normal, less than 25, overweight, 25 to 30 or obese, greater than 30 and then look at people
with brown fat in brown and without in blue,
what you can see is that for each BMI strata, people with brown fat have
significantly reduced odds of the pathology involved. And interestingly, if you look at people who
are obese with brown fat, their prevalence of type
two diabetes is similar to people who are
overweight or normal weight without brown fat. We also can link brown fat
status through laboratory values, and I'll just show you a couple here. So people with brown fat,
again, matched for all of these covariates have
significantly lower, fasting blood glucose. When you plot blood glucose
as a function of BMI, you can see that without brown fat, there's a fairly linear
relationship and that is flattened or blunted in those with brown fat, and we see a similar
relationship for triglycerides. So, what Tobias next did
in a follow up paper was to try and better understand how. And so to do that, he got a little bit more
granular with these scans and because of that granularity,
which required a lot of time actually manually reading scans, we only were able to do this
on a subset of scans done over one calendar year. So this is a few thousand scans. And what he did in collaboration
with Andreas Woodmer was to first look at the CT scan
and by looking at slices from the CT scan, we
could quantify the amount of either subcutaneous
or visceral white fat. And the reason why we did that is because visceral white fat
is particularly pathogenic, whereas subcutaneous, white fat is comparatively
less deleterious, and we wondered whether people with brown fat have a different type of white fat distribution. And then he also went beyond
just this binary categorization of yes or no brown fat to
actually quantify brown fat to see if we could find a
quantitative relationship between brown fat and some
of these other variables. And I just want to show
you two pieces of data from this study. So first of all, we see that people with brown fat have a healthier
white fat distribution. And what I mean by that
is if you control for BMI, people with brown fat have
relatively less visceral fat and more subcutaneous fat. And also now, if you do a
regression analysis where you look at the amount of visceral
fat as a function of brown fat activity, you can see that people with increasing, brown fat activity have
successively lower amounts of visceral white fat. We also looked at the CT scans
to try and assess liver fat. So in prior work, it's been shown that the
liver density is measured by Hounsfield units, as correlated
with liver lipid storage, such that higher liver
density is associated with lower liver fat. And we saw, again, matched for all of the
covariates that people with brown fat have a
higher liver density, meaning lower ectopic
lipid storage in the liver. So in the next, I guess 20
or so minutes remaining, I want to talk about our work
which has really been focused in understanding what is responsible for these benefits
associated with brown fat. And if we come back to our approach, I'll tell you about three stories
that have to do with form. So first of all, how does
thermogenic fat work? Secondly, function, how does it communicate
with other tissues? And third regulation, what are the genetic determinants of thermogenic fat quantity and activity? So in terms of the form, this is work that was led by Jingyi Chi, a former PhD student who's
now a postdoc in Boston, as well as Rico, a current
graduate student in Will Barr, who's actually in the audience
and is now a graduate student at MIT. And they were interested in understanding what are the cellular and molecular components
that can fold the formation and function of thermogenic fat. So just to give you a little
bit more introduction, because we're doing these
studies now on mouse models, in mice, there are three main fat depots. There's the developmentally preformed, interscapular brown fat
that consists largely of brown adipocytes. As I mentioned earlier, these cells have a high
number of mitochondria and dissipate energy
through thermogenesis. Visceral fat is largely white
adipocytes and stores energy, but subcutaneous fat, which
was the focus of our studies, contains a mixture of white
adipocytes and beige adipocytes that in response to cold, can turn on this thermogenic program
and dissipate energy. And so at the onset, we already knew quite a
bit about beige fat cells, mainly through studies either
in vitro or in animal models. So first of all, activation,
most classically, a cold occurs through cold stimulation through the sympathetic nervous system. When these cells are activated, they're marked by the expression
of this protein UCP one, which is crucial for
uncoupled respiration, and this is dependent on PRDM16. And when you activate these cells, this is associated with
metabolic benefits, including reduced body weight,
improved insulin sensitivity, and improved glucose homeostasis. At the beginning of our studies though, there was also a lot that we did not know and in particular we didn't
have any sense at the level of tissue. How do these beige adipocytes interact with other tissue components? And one of the main limitations,
which is highlighted here, has to do with how do we
look at how cells relate with one another in a tissue? And so in traditional approaches,
that's done histologically in these two dimensional sections, either with paraffin embedded sections or with immunofluorescence. And that presents a particular challenge because of its high in lipid content, which makes it challenging to
obtain consistent sections, and then the lipid can
also cause light scatter. And so, what Jingyi did working
in collaboration with people from Mark Hestalevine's lab
is to develop a protocol that we call Adipo-Clear, which allows us to visualize
three dimensional structures in adipose tissues with
molecular resolution. So the starting point is
that tissues are cleared or rendered transparent to light. This is a fat pad before
and after clearing, and then they can be
visualized after staining with antibodies by light
sheet fluorescent microscopy. And I'll talk to you
about some of her work, but this has also been
really useful in a number of collaborations that I highlight here. And so to start, we ask where
are beige adipocytes located. And so this is the subcutaneous
fat pattern of mouse. And we, for naming convention, we call this portion the inguinal portion. There's a large lymph node here and then on the other side of the lymph node is
the dorsolumbar portion. And generally in the past, people didn't consider
regionality of the tissue. So if they looked at the
activation of these cells, they would take an animal, expose it to cold for
varying lengths of time and then either measure
the level of transcripts or proteins involved in the process. And you can see at both
the mRNA and protein level, there's a really robust
up-regulation of UCP one with cold exposure. But when we did whole tissue imaging, and I'm showing you now
UCP one and magenta, after 48 hours or one week of cold, you can see that there's
pretty striking regionality. So after 48 hours, most of
the UCP one positive cells are in the inguinal portion of the tissue. And then after one week,
more cells emerge here, they also begin to emerge in this aspect of the dorsolumbar portion of the tissue. So why is that? Does it have to do with the environment where these cells live or are
there intrinsic differences in the fat cells in these two regions? And we actually think that both
may be true, but for today, I'll focus on the environment
and in particular, the role of the sympathetic nervous system. So, the way that the sympathetic, nervous system activates these
cells is highlighted here. So cold is sensed in the
periphery by sensory neurons. That signal is conveyed to the brain. That then leads to an efferent output through the sympathetic nervous system, first in preganglionic neurons, and then in postganglionic neurons which have their cell bodies in ganglia along the spinal cord, some distance away from the target cell, they send these long
range axonal projections into the tissue parenchyma where
they release catacholamines that binds to metabotropic
receptors in fat cells. That leads to increased cyclic AMP and a signal transduction
cascade, the end result of which is the up-regulation
of this thermogenic program. And so here now I'm showing
you imaging staining for TH or tyrosine hydroxylase, a
marker of sympathetic neurons, and what you can see is
this really rich network of pH positive neurons. You can see these large fibers
traveling along blood vessels where they regulate
vasoactivity as well as axons, but if you zoom in, you
can see these finer puncta that enter the tissue parenchyma. And if you really zoom in, you can see that these are
actually in close apposition to fat cells. And so, we can also
take these clear tissues and do confocal imaging
to get better images of thicker sections and when we do that, what you can see is this
network of parenchymal nerves. And if you look, you can see what looks like
these beads on a string pattern or the varicosities, and
these actually represent sites of neurotransmitter
aggregation where they're then released en masse and
so that the cells closest to them get the highest
concentration of catacholamines. And when we do this, we can
also trace these projections to actually quantify their density. And so when we use these tools and we now look at the
different regions of the tissue, region one is this region
that has pH fat cells. Region three has fewer. And if you look in these
fly through images, you can see that there's
much greater density of pH and green in the inguinal region than in the dorsolumbar region. And when we quantify that,
by tracing these projections, you can see that there's about
a two to three fold increase in neurite density in the
inguinal portion of the tissue. So we next ask, well, how are
these two processes related? And so to address that, what we did was we looked
at either control animals or animals lacking beige fat cells because they have an
adipocyte specific knockout of this E regulator, PRDM16, and what we saw was quite striking. So if you look in control animals, these are just single cut
images in the control animals. There's very dense pH staining
in this inguinal region and very scant staining
in the dorsolumbar region. But in mice lacking beige fat cells, there's very scant nerve density in both regions of the tissue. And that was really unexpected
because what that tells us is that simply by deleting a
key transcriptional regulator in fat cells, we not only affect the
phenotype of the fat cells, but we somehow affect their innervation. And so, this just
summarizes our working model from this work. So it was well known that sympathetic nerves
activate beige adipocytes, but what we found is that these beige adipocytes
are actually required to reinforce their innervation. And if you delete PRDM16, you not only lose beige adipocytes, but you also lose this innervation. So that of course leads to
this chicken and egg question of what's causing what, and in work I don't have
time to get into today, we focus on whether or not this pattern of innervation is
developmentally determined or does it show some
plasticity in adulthood. And if it does show plasticity, how is it affected by things
like obesity, cold exposure, and aging? So in the follow up paper, Jingyi was able to show
that there is this critical, early developmental
period when the patterning of these sympathetic
neurons is established. And at least in our paradigm
where we use inducible deletion of PRDM16, in adulthood, this interaction between beige adipocytes and the nerves is not required to maintain the structural maintenance. And so again, this suggests
that this is regulated by PRDM 16 dependent
factors in the fat cell, and we're now intensely focused on understanding how that works. And so, three possible modes of action, it might be that thermogenic
fat cells release factors that attract nerves, conversely, white fat cells may release
factors that repel nerves, or there could be indirect interactions between fat cells, immune cells, and the tissue
microenvironment and the nerves that dictate their pattern of innervation. And so we've profiled different regions, both by transcriptomics, we hope soon to do proteomic profiling, and we've identified a number
of candidate factors involved in axon guidance, and we are now doing functional
analyses of these factors. So let me now turn away from
form and towards function and understanding how
thermogenic fat communicates with other issues. This has been really a theme effort. It was initiated by Chan Hee Choi, a former MB PhD student who's
now finishing medical school. You can see Will Barr again
features prominently here, and some of the newer work
is being led by Kaja postdoc along with Corey and Samir. And so, you might ask to start, well, why should I even think
that thermogenic fat signals to other tissues? What is the rationale for
saying it has an endocrine role? There are a variety of studies,
none of which prove this, but in my mind, this is
the most convincing study. This comes from Dave
Piston's lab at Washington U in St. Louis. What they did here was
they took a mouse model of type one diabetes. So these mice were treated
with streptozotocin to ablate their pancreatic pilots, leaving them with diabetes with blood glucose values above 400. They then transplanted just
a small amount of brown fat, about 50 milligrams of tissue, one would think far too
little to consume all of that excess glucose
and when they did that, they saw this very durable correction of the type one diabetic phenotype that lasted out to six months. And this remains in the absence of insulin so these animals still
do not produce insulin, but somehow the transplanted
brown fat, it's sufficient, nearly normalized their blood glucose. And so it turns out
actually that fat cells, if you look bio-informatically, are predicted to secrete more than a thousand unique polypeptides and probably an even greater number of small molecule
metabolites and peptides. And the vast majority of these molecules, as well as their targets and functions, have not been characterized. And one of the major stumbling
blocks has been a lack of suitable technologies,
especially in the proteomic space. And of course, if we
can solve this problem, we have the potential to
answer big questions such as the length between obesity and disease, how brown fat might protect
and perhaps even lead to new therapeutic targets. And so for proteomic work,
the major challenge of course, is that proteins in the blood
exist over more than 10 orders of magnitude in concentration. And so, if one does traditional
mass spec of plasma, you will detect a lot
of albumin and globulin but it's very difficult to
identify proteins in the nanogram or epigram for middle range, which is where we think
these hormones exist. And so, we've used a variety
of chemical approaches. I will talk about one today called BONCAT or bio-orthogonal non-canonical
amino acid tagging. Our initial studies, which I'll tell you
about, were done in vitro, but we are now doing this in a vivo and I'll explain how that works as well. So the way that this
approach works, of course, normally in protein synthesis,
methionine, for example, is incorporating into the proteome by a methionine and tRNA synthase. What you can do is replace methionine with a methionine analog, in this case, called azidohomoalanine or AHA. And this analog contains an azide group which does not exist in
amino acids in nature. And why this is useful is
because from this point on, all newly synthesized proteins
will be azide labeled. And the reason why we care about that is because we can now do an
experiment like this one where we can take primary fat cells from either the brown fat
depot, the subcutaneous depot, or visceral fat depot, replace the media with the
media containing AHA instead of methionine, keeping the serum there. And so, all of the newly synthesized
proteins will be azide labeled. We can collect those
from the condition media and then use quick chemistry
where we take a bead coupled to an alkyne mohyeldin which will form a covalent
bond with this azide group, use the bead to pull these
proteins out of the mixture and then digest them into mass effect. And so, that's exactly what we did. And when we did that with
these three different kinds of adipocytes, we identified more than 600 unique proteins
secreted into the media, and about 350 of them showed
enrichment in the secretome of different kinds of adipocytes, which you can see in this heat map here. So here is a collection of proteins that are secreted particularly
by brown fat cells. Here are proteins secreted
by visceral white fat cells. And if you look at this
functional characterization, you can see interestingly, given what I was telling
you in the previous story, that the proteins secreted by
brown fat cells are enriched for molecules involved in axon guidance and neuron projection guidance
when we do go analysis. And so we, of course, don't
just want to assemble lists of proteins but we want to
identify proteins and do biology on these proteins. And so, Chan Hee's work, which was actually accepted
for publication yesterday, focuses on a white fat derived protein that he shows can promote
insulin sensitization, but we've also identified a number of other interesting candidates, including a cold induced
secreted protein made by brown and beige adipocytes that's
almost completely unstudied and the function of which
continues to elude and vex us. So I hope to solve this problem one day in the not too distant future. We're also moving towards
doing this in vivo and the reason for that,
I think, is obvious but it's not possible to
adequately model a variety of pathophysiologic conditions
in vitro like obesity, fasting, and cold exposure. And then secondly, as good as
these cellular systems are, they are heterogeneous
and somewhat artificial. And lastly, we don't want
to just identify proteins that can be secreted by a cell, but proteins that actually
circulate in the blood at physiologically meaningful levels. And so it's now possible
to do that with a variant of the approach that I told you where now a different methionine analog, as you called, azidonorleucine or ANL. You'll notice for those of you
who are chemically inclined that this has a more
bulky side chain group. And so as a result, this is normally not well incorporated into the nascent proteome by wild type methionine tRNA synthase. However, in this work done
by Aaron Schumann's lab, what they showed is that
making a single point mutation in the methionine tRNA synthase, an L to a G now allows
it to incorporate ANL into the nascent proteome. That is very useful because they have made a
conditional mouse model, which they were kind
enough to share with us, that has knocked into
the rows of 26 locus. A lock stop locks in front of this mutant methionine tRNA synthase. This allows you to then cross
it, your pre-line of interest, and then when you introduce ANL coupled to whatever physiological
state you're interested in, only proteins made in cells expressing
pre-recombinase will be labeled. And so we can again now use this approach to enrich these proteins from the blood and we believe that this enrichment approach
offers us the potential to identify low abundance proteins that would typically elude us
with conventional mass spec. And in principle, this chemical handle could
even allow us to track proteins in their destination. So this just shows you a proof
of concept of this approach. So if we use this mouse strand
cross UCP one increase so that we label only UCP one positive cells and house mice either at 30
degrees or at eight degrees, you can see the pre-negative
mice give almost no signal so very low background, and we can see very clear
the labeling of proteins in the pre-positive animals at
30 degrees and eight degrees. And you'll have to take my word for it but if you look closely, there is some unique banding pattern of these two different temperatures and we're now beginning to try and uncover what these proteins are that are actually secreted. We're also taking a variety
of broader approaches. So I mentioned BONCAT. We are also using
another chemical approach in collaboration with Alice Ting, (indistinct) and others
using proximity labeling. This is also a pre-dependent
approach in which proteins in the ER pathway are
biotinylated in a cell type of interest. And again, we can use this biotin handle to enrich the proteins in the blood. We're also doing bulk
analyses of serum and plasma, both looking for proteins as
well as lipids and metabolites and then we've also developed
an increasing program, doing studies in humans, and I want to take a few
minutes to tell you about that. So because this all originated in humans, I feel it's particularly
important to study this problem in human subjects, and so
we've developed two studies. The first is well underway. We call it the cold vest study and this looks at changes
in plasma proteins, metabolites and lipids in
response to acute cold exposure. And the second, which
just received IRB approval and we hope to start soon, looks at the effects of
chronic cold exposure. Let me tell you a little
bit more about this. For the cold vest study, we recruited men and women, normal BMI between 18 and 28 years of age. We picked these conditions
because we specifically wanted to look at people who had
a very high likelihood of having cold induced brown fat. Because of logistical
challenges and costs, it's not possible to
screen each of these people with a PET scan pre and post, but we excluded people
with known pathology, any prescription medications, illicit drugs that might contribute. And so, what we did was
we recruited these people. They came fasted. We did some pH phenotyping
and measured body composition. We then took a baseline blood
sample and then exposed people to cold for a three hour period, using a cooling vest protocol. This is a well established
method that was developed by Aaron Cypis and others at
the NIH where you put on a, it almost looks like a life jacket that circulates cold water, and the idea is that you
customize this to each individual. I think we all know that different people have
different cold tolerances. And so, we progressively
lower the temperature until the subject begins shivering, and we call that the shivering threshold and then we raise the
temperature two degrees and that's where the patient
sits for the remainder of the study. And then at the three hour
mark, we collect blood again. So we've done these paired
studies on 20 subjects, and we've now done a variety
of assays in collaboration with Rob Carissan and Allen Cycathelium, including SOMAscan and
O-link to look at proteins, metabolomics and lipidomics at the plasma. But the goal of identifying factors that are regulated by acute cold exposure, possibly one day leaving
us towards biomarkers and also these might be
functionally important candidates that we can assess further in
cellular and animal models. So I just want to show you a
bit of data from this work. This is all actively underway. This is from metabolomics analysis and there, it's about 400
or so metabolites shown here in this volcano plot. In blue are things that are
up-regulated in the cold. Red are things that are down-regulated. And I just highlight too
as a proof of concept, so DiHome is a lifted
mediator that's been studied by UA science lab at the
Goslin and shown to be induced in cold in both animals and humans, and to also facilitate fatty acid uptake by the skeletal muscle. And what you can see in our
subject is this is regulated in response to cold exposure. We also see BMGV. Don't ask me to tell
you what this stands for because I'm drawing a blank, but this was previously
identified by Rob Gerslin's lab as being a marker of hepatic steatosis and risk for development of diabetes, and you can see that the
levels of BMGV go down in the blood with cold exposure. These are two of many
interesting candidates that we're increasingly uncovering. The second study that we hope to soon kick off is called the MIWI Study, or Minneapolis Ice Water Immersion Cohort. And you may find this a bit crazy. This actually started by a group in Minneapolis contacting us to ask if they could participate
in research on brown fat. And so there's actually a group of several hundred individuals
who voluntarily cut holes in the ice and swim in these frozen lakes during these frigid winter months, and they feel that it provides them with tremendous benefits, both
physical and psychological. And so we thought, well, maybe there's actually
something interesting here. And so, what we've done
is developed this study where we will study these individuals both in the winter when they're
taking these cold swims. Some of them do it every day, and then in the summer when they're not. And so for all of the individuals, we will profile their plasma
proteome, the pabolome and libidome and for a subset of subjects. We'll also do deeper phenotyping
including hyperinsulinemic to glycemic clamp studies,
FDB PET scan studies, and we'll also obtain fat biopsies to do single nuclear sequencing
to look at the cellular and molecular landscape
of their adipose tissue. So, this was actually a long
slog to get approval for this because it involves two institutions since we're not where these patients are. And so, we're excited to be
collaborating with Betsy Sequis at the University of Minnesota and hope to kick this off very soon. So in the last, I think I
have just a couple minutes, I want to turn to the regulation. So ultimately, I think if we're going to
get anything actionable here, it's important that we understand
what regulates brown fat in humans. And looking again at this
variation in brown fat quantity and activity, it begs the question, what are the genetic determinants of this that hasn't really been studied before? We know that obesity is a
highly genetic phenotype. Might brown fat be under
genetic inheritance as well? And so, I just want to show
you data from two other studies that at least suggest
this isn't a crazy thing to think about. So this is from a retrospective
study from Aaron Cypress and Ron Con where they
looked at brown fat quantity in men and women. You can see women have
more brown fat than men, as I told you. But if you look at the scatter plot, what you see is there are a small number of individuals who have
exceptionally large amounts of brown fat. So in other cases, these extreme or exceptional
variants can sometimes be used to identify aminos with large effect size that contribute to the phenotype. This is from a separate
study done in the Netherlands where people did PET scans
to measure brown fat volume of either Caucasian subjects
or South Asian subjects living in the Netherlands. So it was meant to
control for environment. And what you can see is that the South Asian subjects
have a significant reduction in brown fat volume, at least
consistent with this being under some form of genetic inheritance. So we're now taking several
different approaches and I'd welcome any suggestion. I know there's a number of
geneticists in the audience, and perhaps on the Zoom call, I'll just briefly
mention what we're doing. So the first is that we've gone to our Sloan Kettering cohort
of about 50,000 patients. This was a retrospective study, but fortunately a number of the patients in our study have had
their exome sequenced as parts of other studies. And so, we were able to piggyback on that and we've specifically focused on people who either have exceptionally
high brown fat activity, greater than the 90th
percentile for their age, or who have what we call
age retained brown fat. As brown fat prevalence
goes down with age, we reason that people over 50 who still have brown
fat must be rather rare because the prevalence is
like, less than one percent. And so we have a cohort of about a hundred people
who fit those criteria, and we're comparing their
exome to reference exomes and have a number of candidates that we're beginning to study. But we're also looking at two different well defined cohorts. The first is the Turkish obesity study in collaboration with Tayfun Oxoelik at Bill Kent University where he has done exome sequencing on about a thousand people
of BMI greater than 40 from consanguineous pedigrees,
and the hypothesis here is that these individuals
might have mutations that would give them
decreased brown fat activity, therefore decreased energy expenditure and predisposition to obesity. And then finally, the third cohort is in
collaboration with Sadaf Farooqi in the UK. She has a cohort of people
who have a thin phenotype. This is not commonly done, but rather than sequencing
people who are obese, she has sequenced the exomes of a few thousand people who
are thin with a very low BMI but have no underlying
medical pathology that seems to explain it and the prediction is that these people might have variants that give them high brown fat activity, increased energy
expenditure and protection from weight gain. And so the paradigm
that we're following is to identify these
variants and engineer them into cell lines to phenotyping
both at a molecular and bioenergetic level and
then for promising candidates, make animal models when the
residues are concerned to see if the phenotypes match. And so just to summarize, with this backdrop of obesity in all of its dependent medical complications, we believe there's some
reason for optimism in that brown fat, seems
associated with protection from cardiovascular
and metabolic diseases. And so, we've really focused now on understanding how that works. I showed you data about
the form of thermogenic fat and how it can be regulated by crosstalk with the sympathetic nervous system. I've also told you about
our work on the function, on the regulation rather,
and the endocrine action of these issues. And so the long term goal, of course, is to use all of that information to try and develop therapeutics to break the link between obesity and associated diseases. And so in closing, I just want to thank all
of the people in my lab, our collaborators, and I thank
you all for your attention. I'm happy to take questions. I see there are questions in the chat. - [Anna] I was going to say, Paul, the folks online have been very patient. So first, Ed Schulmick wants to know, is there a way to activate
PRDM16 with small molecules and not require the sympathetic
system to activate it? - Yeah, so great question. I mentioned some of the students
who I had lunch with today, when I was a postdoc, I spent a few months doing
a small molecule screen at the Broad looking for
small molecule activators of PRDM16. That work unfortunately was a failure, not because the screen failed, but because we were too
naive in our understanding of how PRDM16 was being regulated. So at the time, we assumed it was regulated
transcriptionally. And so we did an RNA based
screen and we identified only two or three hits, none of
which really bore out. It's now become clear
over like, the last decade that PRDM16 is largely
regulated post-translationally. And in a recent paper, In
Nature from Shingo Kajimura, he identified the mechanisms
that regulate the turnover of PRDM16. So I think those, (indistinct) that regulate the half life and turnover of PRDM16 are probably
especially appealing targets for ramping up on that emphatically. - [Hernandez] You can hear me? - Yeah.
- [Hernandez] Great. Amazing talk I really love, and I every time that I see your work, I get really super excited. I want to follow up this question
of ED and also connect this with two recent works that
is coming out from Lee E from scripts and also from (indistinct) about the neuropathic
when you have obesity, but you have neuropathic, you lose sympathetic system in the fat. And linked together with
this PRDM16 phenotype, that you see that when you knock out, you have less innervation. So by what you guys see in
terms of gene expression in the beige adipocytes for development, do you see loss of any
neuropeptide or any growth factor that could be mediating
innervation or if you see even with PRDM16 knockout, a lot of innervation
early on in development and how this could play a role in that? - Yeah, thank you. Thank
you for your question. So what Hernandez was alluding
to is work from others in the field showing an
important role of sensory nerves, first of all, in regulating these cells. I focused on sympathetic nerves and also on the role of left and regulating sympathetic
nerves and how obesity may or may not alter that. So in our studies of the PRDM16 model, we used also a docs inducible system. I didn't have time to get into this. And we saw that if we delete
the PRDM16 and adipocytes in this crucial window early in life, then we lose the innervation and that's not rescued later on. But if we delete PRDM16 and adipocytes in fully mature adult animals, we don't see any apparent
effect on the innervation. So that has then led us to
focus much more intensively on what's going on in
this crucial early period. We've identified a number
of candidates and that focus on the early period has also led us into some interesting insights of where the beige adipocytes
might actually come from. I hope to have more on all of that soon. - [Speaker] So it's fantastic work. I was really curious about the vast study. And you regulate by how much
cold the individual feel, and is there a correlation with already having more brown
fat or beige fat to feel more and then this can be a
viable, you know, study? - Yeah, that's a great point. One of the major weaknesses of our study, and I hope you don't review it one day because you've just highlighted it, is that we don't have PET
scan data on our patients. So our research hospital is
great for doing small studies. We don't have a PET scanner research. PET scans cost a couple thousand dollars. And so what we would really
want to have is both a baseline and a cold induced PET
scan on all of the subjects in our study. We don't have that, however we are going to be
collaborating with Aaron Cypis, who's at the NIH who has
one of the greatest sets of such samples. And so, the plan will be to
look at blood from people with well defined brown fat
quantity and activity pre and post cold and do the same assays both to validate our findings, but also to look at things
that are maybe correlated with brown fat quantity and activity. - [Speaker 2] More virtual
question or can I go ahead? Yeah, wonderful talk,
wonderful work as always Paul. So I have a few questions. I'm just going to be really quick and I'm glad we have more
time to discuss later. One is for the sympathic
nervous system interaction, which of course makes complete sense and has been modeled in the dish before. There are studies from Bruce Fegelman and others showing just simply
reducing the temperature of the incubator, inducing the phenotype
of beige in both sides. Could you comment a little bit on that? - Sure. Yeah. So that worked from the same media that Hernandez alluded to when
he was a graduate student. So this was a study where
former students took fat cells in a dish, so no nerve, no blood vessels, and he lowered the
temperature of the incubator and showed that the thermogenic
program was up-regulated and when he then raised the temperature, it was down-regulated. And so what that proves is that these fat cells
must have some intrinsic, cold sensing mechanism. What exactly that
mechanism is is not clear, but there has also been really
interesting work recently from Alex Mither and Bon, which has highlighted sympathetic
independent mechanisms for activating these cells, either through adenosine or inosine. So I don't mean to imply that sympathetic nerves are the only way to activate themselves, but that's at least the
most well studied way. - [Speaker 2] Yeah, great.
Quick other question. So for your, you know,
genetic variation study, are you focusing only on
protein coding rare variation with the exome sequencing? Because as you know, we have
worked in the, you know, common coding variation space where I guess (indistinct)
work and our work provides at least some evidence that non-coding genetic variation
modulates lesion capacity of adipocytes. Is that of interest at all? - Yeah, absolutely. I met with Jose after I met with you and he asked me to say it. So we decided we had to start somewhere and because we have more
subject with exome data than old genome sequencing data, we thought we would start there. We also have perhaps some genotyping data so we could have some
incomplete, old genome data. But to start, we decided to focus on
the exomes with the hope that we'll find rare alleles
with large effect size, but we certainly don't
intend to neglect looking at non-coding variants as well and we value your expertise
when we get to that. - [Speaker 2] Thank you. - [Speaker 3] Yeah, great talk. I had two questions, one specific one, a little more general. The first one is, do you predict that if
you were to express PRDM16 in say like the lumbar region, would you shift the innervation
pattern to innervate that region more? - Yeah, so great question. I think we would predict
that if it's expressed in this crucial developmental
window, perhaps yes, but we don't know for sure and the reason we don't
know for sure is because in these imaging studies I showed you, we haven't been able to
do staining for PRDM16. And so what we'd really like to see is, is PRDM 16 enriched in this
highly innervated region or not? We think it may be, but we'd
like to be able to prove that and undoubtedly there are
going to be other molecules that interact with that
are important as well. So I don't know if the one protein alone
would be sufficient. - [Speaker 3] Thanks, and my
second question is, you know, you showed that there's a
lot of genetic variability with how much brown fat
they have at baseline. Do you anticipate that there
will be similar variability in how much fat is
inducible within people, or is there sort of like a ceiling at which everybody would get at? - Yeah, so great question. I
think we don't really know. I'd like to think that both
are under genetic control, but the quantitative
data is harder to assess in large cohorts because
in the clinical setting, we benefited from the fact
that when the scans were read, the reader said brown
fat was detected or not. And so, we could do that
binary categorization. If we want to actually quantify
brown fat volume or activity, someone has to manually review the scan. It would take about 20 minutes, and so to do that for thousands of fans is an enormous undertaking. I'm hoping that it will soon
be possible with automated or semi-automated approaches to do that, but we're not doing that. - [Anna] Three more questions. So just curious, going back to
the first part of your talk, since you had such a large cohort, did you analyze any
factors such as exercise that may be associated with
the brown fat distribution? - Yeah. Great question. So because we had a retrospective cohort, we could only use the
information we had available in the charts. We focused on diagnosis codes because those are readily available and as any physician knows, because they're the
bane of your existence, you have to do these after each visit. So we don't have clear, granular
data on physical exercise, but we would like to look at that. There was a paper recently
published that just came out, though, showing that an
exercise paradigm in humans, this was a prospective
study, was not associated with the activation of brown fat. So it could be that that's one area where the rodent model
and the humans differ. - [Anna] Then is there any suggestion that brown fat protects
against degenerative diseases? - Was it neurodegenerative or unclear? Yeah, so we are not working on that, but there may well be. There's a colleague at Augusta
University, Alexis Stranahan, who's published some really
interesting work on links between thermogenic fat and
Alzheimer's in mouse models and she shows that there can be changes in cognitive performance
that presumably are due to some fat brain boss talk
that her group's working on characterizing. - [Anna] Final question. Great talk. What do you think about
the relative importance of thermogenic contributions
from skeletal muscle versus brown adipose tissue? - Yeah, so I guess the point
is that when one is cold, the normal response is to shiver and that also generates heat. That's why we do it. And so, if we're talking
about heat generation, I think that undoubtedly shivering
contributes more to heat. Our massive skeletal muscle is way, way higher than our mass brown fat. And so, our prediction is that the benefits we see associated with brown fat are more likely
to be mediated by signals to other fishers than by its
energetic contribution than so. - [Anna] Fantastic. There weren't questions, but
I think we have to end here. This was great. This was actually at the peak of it, I think, 120 or so participants. So we have great participation
between present and online. Paul, I can't thank you
enough for a fantastic seminar and thank you all for coming. See you next time.