John J. O´Shea

John J.  O´Shea


Thanks so much for inviting me to this very prestigious
lectureship. I was thinking, you know, when I got invited, you know, I wanted to
come up with a talk that would really impress you. And I then I thought, “What a
ridiculous idea. Here I am in a city that has had, you know, lectures describing the
most important discoveries of mankind over the last century or so,” and I
thought, “I’m not going to say anything that’s going to impress you.” So in
thinking about that I thought, “Well, what could I do?” And I thought, “Actually, I
think I’m just gonna tell you a story about what I’ve been interested in, and
then you can decide whether that’s an interesting story or not.” And as Ron was
saying, I wanted to start off with where I started at the NIH and … and what
seemed really interesting to me at the time and where that sort of led me in my
career at the NIH and how I was thinking about, as I’ll show you, in the early days I
was really intrigued by the concept of receptors and how they cause cells to
change their behavior. And in the course of that, as Ron alluded to, I’ll talk a little
bit about, and then I’ll explain how that, a very basic idea, led to a new class of
drugs, but equally how interrogating patients leads you backwards to think
about these things again and then have… what the other thing that happened, of
course, during the duration of doing these experiments is, well we completed
the human genome for one, and now we have all sorts of genomic tools that you guys just take for granted. But, frankly, there are things that we couldn’t even conceive of
when I started thinking about this. So I think, as you’ll see, we’ve filled in a
lot about this black box over the last 25 years or so. But I think equally
you’ll see that we’re just at the beginning of understanding these processes. So when I was an undergraduate I… I took a course, and it was about the time when this
paper came out, and they had made this spectacular discovery of how receptors
sat in the plasma membrane. And they were like these, you know, little potatoes
in this… in this soup, you know, and… but importantly, what they did recognize was,
you know, how this might happen physically and chemically. At the same
time, as intriguing as this was, it sure didn’t tell you a lot about how a signal
out here could change a cell’s behavior. So I came to the NIH, as you heard,
clinically trained and, as the students… we talked about at lunch… that I really
did get the bug for understanding basic processes. And as you heard, mostly
because I had been a clinician, and I thought about options, and doing research seemed
way more interesting than some of the other options. And really the NIH was the
right place to go for a young clinician, but I didn’t realize right away I was at
a disadvantage because I didn’t have a Ph.D. So I knew medicine but there was a whole lot that I didn’t know, and I certainly didn’t know much about this
process of receptors and receptor-mediated signal transduction.
But I was fortunate enough to take a course at the NIH on biochemistry and
signal transduction, which, ironically, I now teach in, but this was
one of the professors who was teaching this course. And he gave this amazing
lecture on G protein-coupled signaling and I remember at the time I was just
spellbound. I thought, “Wow!” I mean I guess I sorta had been sleeping
during that part of the class in medical school, or maybe they hadn’t
discovered it yet, probably more the latter, and I remember when I went up to him and I said, “That was a great talk!” you know, and he said, “Thank you.” And then shortly thereafter he did win the
Nobel Prize. So he was probably really waiting for me to thank him and say how great a talk it was. But I was thinking about this when I was putting my
talk together. I thought, “That’s really so exciting now to be coming back now.” And
then to again be thinking about how… I guess he got the Nobel Prize in 1995 or
so… and just think about how this problem has changed since I heard these lectures
in the 1980s. And so what was happening at that time is that, of course,
this was the molecular biology revolution, so we went from going…
thinking about these potatoes swimming in this lipid soup… to really having
structural insights into receptors and so again I remember very vividly at one
of the meetings I went to when the insulin receptor was first cloned and…a
very exciting discovery… was that the insulin receptor was a receptor tyrosine
kinase. So for all the, you know, struggling of trying to understand how
such a receptor could work, suddenly it was easy. The receptor was a kinase…is a kinase… and
you could imagine if insulin binds it activates the enzymatic activity and you
induce phosphorylation of substrates. So that was tremendously exciting. What was
also exciting, of course, was when you added cytokines to cells that also induced
tyrosine phosphorylation, but actually the cloning of tyrosine… of cytokine
receptors… was a little bit of a disappointment. One of the first cytokine receptors that
was cloned by my colleague Warren Leonard at the NIH…he cloned the alpha chain of the IL-2 receptor, and again that doesn’t even have this configuration in it and
really has a very small nubbin of a cytoplasmic domain. So suddenly
that answered a question but it raised many other questions. And really the question was on “What’s
the kinase that’s responsible for signaling via cytokines in this class
of receptors?” So it was solved, as you well know, by the discovery of the JAK STAT
pathway by Darnell, and George Stark, and Ian Kerr, and others, and it
came from two approaches really. One was just screening libraries for kinases,
but then also a pivotal experiment by Sandra Pelegrini who showed that if you took cells that weren’t responding to a cytokine, in this case
interferons, and reconstituted them genetically, you restored the pathway. So that really put JAKs… in this case it was just to make things complicated for you that
JAK was called Tyk2, but as I’ll show you in a moment, Tyk2 is a Janase family kinase… but that restored signaling by interferons. And Jim Darnell, of course, took a
more sort of brute force approach by identifying complexes bound to
interferon-induced genes … discovered STATs which we now have seven STATs. But what we now know, of course, is that this is a pathway that’s employed by more than
60 cytokines and growth factors. It’s evolutionarily conserved really easily
from flies to mammals with the pathway that is very similar. Dictyostelium to mammals also have elements of this pathway, and I think Dictyostelium remarkably has four STATs, so this is a
pathway, right, when you think of Dictyostelium, it’s this organism that
can’t quite decide whether it’s a single…a unicellular organism or a multicellular
organism… and it sorta changes when things get bad; they decide to band
together. And so this obviously is a very very ancient pathway that controls an
organism’s response to the environment, and what we know now, of course, is that
this pathway is critically important for development, growth, hematopoiesis,
metabolism and immunity. The question, of course, is that what
hadn’t been sorted out, really, was the in vivo criticality. So in these cell lines,
mutantinized cell lines, we knew that this pathway was important, but again this was
at an era where people then moved from experiments in cell lines in this sort
of thing to looking in vivo and using knockouts. Now in this case, the first
knockout in the JAK STAT pathway didn’t really come from a mouse. It came from studying humans. So Gigi Notarangalo and Rebecca Buckley had patients who had
a very particular type of immunodeficiency, and what was discovered, again, my colleague Warren Leonard at the NIH, had discovered that mutations of the
common gamma chain, a receptor that’s used by IL-2, 4, 7, 9, 15 and 21, was lacking in boys
that had this particular type of immunodeficiency where they had no T
cells; they had B cells; they had no NK cells… referred to as T minus B-plus NK
minus SCID. When we cloned JAK3, we had shown that it’s activated by things like
IL-2, and then, with Warren, I was able to show that JAK3 and the common
gamma chain were associated. And when this paper came out, people like Gigi Notarangalo and Rebecca Buckley reasoned, appropriately, that… that maybe there were
girls who had T minus B-plus SCID who had not mutations… they didn’t have mutations
on the gamma chain, of course, because it’s X SCID, but autosomal recessive SCID might be due to
a JAK3 mutation. So we collaborated with Gigi and Becky Buckley and… and in
fact reported girls and boys who had SCID due to JAK3 mutations. So
that was important for a couple of reasons. It demonstrated that, in fact,
JAK3 is critical for signaling by the common gamma chain. But it also
was important because this disease really phenocopied this disease, so it really said that the function
of this kinase was for signaling by this receptor; it didn’t have broader
functions. This…mutations in this receptor didn’t have broader functions… that this
really was the pathway. So around this time, of course, there was a lot of
interest in targeting cytokine receptors; again this was the era of, you know, molecular biology explosion where many receptors and many cytokines, etc., were
identified and then, of course, we began to make monoclonal antibodies against
cytokine-cytokine receptors. So when we published this paper, we made the bold
claim, way back in, you know, whatever it was, 1995, that if you targeted JAKs
that would represent a new class of immunosuppressive molecules,
immunomodulatory molecules. So about this time, I was at a meeting in Saxtons
River, Vermont, and just presenting the cloning of JAK3, and I ran into a guy, Paul Changelian, who actually I knew previously… we both worked on
complement prior to what I was working on now…what Paul was working on… and I said, “Paul, you know, what are you doing at this meeting on… I think it was antibodies… and it was an immunology meeting… what are you doing at this meeting?” And he said, “Oh, I’ve now moved to Pfizer, and we’re looking for tyrosine kinases to target.” And I said, “Well,
I got a winner for you.” And so, in fact, for collaboration then with Paul and a chemist,
Mark Flanagan, Pfizer generated this compound here, called Tofacitinib, and this was an example where at the NIH we certainly can collaborate with industry. And this was
done in the form of… it’s called a CRADA which stands for something…I don’t know… “Collaborative Research Agreement and Development Award….” And so really we’ve had since… since that time, for probably a couple of decades, we’ve had collaborative [correcting] a collaboration with Pfizer.
And, again, it’s… for me… it’s been very very very exciting because you sort of
realize right away the strengths of academicians, and then you also recognize
very well the strengths of industry in… in really taking, you know, what’s just sort
of an offhand comment in, you know, a discussion section of a paper to
moving something like that to reality. So this is the current status of Jakinibs, JAK inhibitors. There are the FDA approved JAK inhibitors shown
here. Tofacitinib I will come to in a moment. Ruxolitinib was actually approved by the
FDA first, and again that was sort of a little bit of a funny story that Ruxolitinib is a JAK1/JAK2 inhibitor. It was recognized during the interim that the disease
polycythemia varia myelofibrosis is due to gain of function mutations of JAK2,
and so, in fact, insight was well poised then to develop and study in clinical trials their inhibitor in this disorder. It’s being tested in other disorders as well. So I’m gonna come back to this again. I’ll simply point out
now that at around this time people then made knockouts of all these different mice.
And if you made a knockout of JAK2 what you ended up with was embryonic lethality.
JAK2 is the enzyme that’s important for signaling via erythropoietin. So you
can imagine, as an embryo, if you can’t make blood, that’s not a good thing. So,
and I’ll come back to that, so this whole idea in drug development of using
genetic tools and predicting utility in clinical scenarios from your genetic tools, etc., this I think is one of the sort of interesting
things to sort of to think about, and JAK inhibitors are a particularly useful
lesson in this regard. I’m gonna come back to that. Tofacitinib, again we had a CRADA with Pfizer… this drug is a JAK1/JAK3 inhibitor, and I’ll have to tell you more about that in a moment, and less a JAK2 inhibitor. It’s approved in the United States for rheumatoid arthritis and there’re
ongoing trials in juvenile arthritis spondylarthropathies, IBD, and
psoriasis. So I was telling Ron, and we were at dinner last night, and I was in Amsterdam, I
guess about a year or so ago, and again they invited me to come speak about JAK
inhibitors and again they were very very gracious and they referred to me as the godfather of this class of drugs and everything like this… very flattering… but
at dinner somebody told me a story that’s very similar to the history that
I’m gonna tell you. So I’m gonna tell you about Leo. So Leo is a six-year-old and he was well until about
three years ago when he developed severe dermatitis, and his mother brought him in
to see the doc, and he was diagnosed as having food allergies, and then they did
everything you do with food allergies… you restrict diet and all this complicated stuff, and he wasn’t getting better. And then he was put on steroids, and again his mother was really really worried, you know, putting a six-year-old on steroids, what with all the
side effects of steroids and everything, and still he wasn’t doing well. And this is Leo, right here. So you can see Leo is just miserable. You can see all the
dermatitis. Leo is very unhappy and really had a modest response to prednisone. Then his mother found out about Jakinibs. And Leo was put on Jakinibs and you can
see Leo is a happy dog. And Leo’s mother works at the NIH. Leo’s mother is a
scientist, and ordinarily if you’ve seen me give this talk I think I usually black
out the dog’s eyes for the privacy purposes. But Leo’s mom said very specifically I
could use these slides and was actually thrilled that Leo is going to be shown at
the Nobel forum that… you know…. And Leo was wagging his tail when we discussed this
possibility. So that leads to the third FDA-approved JAK inhibitor, Oclacitinib,
which actually came from the original screen of Pfizer. Pfizer then spun up… this
is a different company now… but it was originally part of Pfizer, and Oclacitinib is
a JAK1 JAK2 JAK3 inhibitor and it’s approved for dermatitis in dogs,
and I can tell you often when I go around and give this talk… I… at the end in the reception, I often have more people asking me about their dog than
patients with all these diseases. I should add, though, I think it’s actually very interesting having this veterinary experience, because now we have many many, not people, but many many subjects, if you will, on a JAK inhibitor, so we’re really rapidly
expanding what these drugs can do, and it was actually interesting, I won’t
go into the details now, but how these trials were done to get FDA approval. Now
in the meanwhile, people are using JAK inhibitors for other things. You can
see that one thing that really attracted everyone’s attention is that JAK
inhibitors can cure baldness. I, needless to say, I got very excited about this, but this is a
particular type of baldness, and this is autoimmune-mediated baldness… several types in which JAK inhibitors have been used. And the drugs have also been used for other
immune-mediated diseases such as vitiligo. So let me just go back and
and just imagine, if you will, when I first went to Pfizer in, you know,
whatever, the mid-nineties, and made the pitch saying, “Oh, no, you guys should target JAKs,” and what we were thinking. So at the time, of course, based on our discoveries on the pathway and then the patients, we thought, “Well, what you really want to do is make a selective JAK3
inhibitor.” Now again, without going off on too much of a tangent, I would just sorta remind you… back then, before we had so many tyrosine kinase inhibitors, believe it or
not, it was actually controversial as to whether tyrosine kinases would be a good
idea or not. You might imagine that if I said, you know, “Well, I think it’s a really
good idea to target an enzyme that uses ATP.” You know, people would look at you and think, “Really? How you gonna get some specificity?” Now we all take that for
granted now… that chemists are so good that they can make selective inhibitors. But
the idea then of targeting an enzyme that used ATP seemed ludicrous. And then… but
if you said, “OK, you know, what we really need to do is target this particular
enzyme,” people were… it was a little controversial, but people were OK with that. Now I point down here that at the time we
thought of IL-2 as target… as JAK3 as being important for IL-2, 4 ,7, 9 and 15. And
again, based on the patients, we were thinking of primarily this drug as targeting adaptive immunity. But, and this is another point that I’d like to sort of
make, is that when a drug development project starts, you have a certain batch of
knowledge, and then what happens? Well then you go ahead and often it takes 20
years to get a drug, and science doesn’t stand still. It changes, right? So you
think about way back in mid-nineties we didn’t know about IL-21, right? We didn’t
know about Foxp3. We didn’t really think about Tregs. TH17 cells hadn’t been
discovered. And Charlie Janeway… actually the first time I sort of learned about
the concept of innate immunity was actually here in Stockholm… Charlie Janeway was at a talk, and he was pitching this idea of innate immunity. And I remember thinking, “What
is he talking about? What is this innate immunity thing? This is… if you’re working on… this
is all nonspecific stuff. This is really where all the action is.” And so we didn’t really even think about
how this drug might be targeting things other than JAK3; that seemed very
logical. And then on top of that, well, the human genome hadn’t been completed, so we didn’t even use the word “kinome,” so there,… you didn’t have the opportunity to think about selectivity on a broad scale. But now, of course, we have the human kinome; there are 518 kinases, 90 tyrosine kinases and four JAKs. Again
when we were just cloning the JAKs, we didn’t really know how many JAKs we would
ultimately have. So now it’s a pretty straightforward thing. If a company
says, you know, “I have a new inhibitor, a new JAK inhibitor,” as a reviewer you just
say, “Oh yeah?… you know… show me, Buddy.” And this is the experiment that they can
readily do. They can take the kinome and then say, “OK, does the drug have some
kind of selectivity for the kinase you want it to have?” You say, “It has actions
against, but more importantly, perhaps, does it have activity against other
kinases in the kinome?” So that was done with Tofacitinib, not by Pfizer, but by other
companies, and we now know that Tofacitinib, for example, the big circles mean
high activity towards that kinase for JAK1, not shown here, JAK2, and JAK3, in vivo it has a little less selectivity… has less activity
towards JAK2, and pretty much not a lot of activity towards anything else… another enzyme down here, that it’s not
clear that that has any relation to efficacy. But, so now we know that JAK3, Toficitinib, inhibits JAK3 but also inhibits JAK1,
and then to a lesser extent JAK2. And I can only imagine if I had gone to
Pfizer, you know, back in the early nineties, and I said, “You know what, why don’t you just make an
inhibitor that activates JAK3, JAK1, JAK2, and that’ll be fine. That’ll be fine if you’re
inhibiting all these JAKs.” They would have said, “Thank you very much. Don’t come back. That’s the stupidest
thing I’ve ever heard of.” But as you see, we’re in a little different era now, and, in fact, these are all the cytokines that are potentially targeted by a drug that
inhibits multiple JAKs. So Tofacitinib will inhibit JAK3, JAK1 and JAK2, and so you’re
inhibiting the thing we expected to inhibit, adaptive immunity, but you’re also
inhibiting cytokines that are really pivotal in innate responses like IL-6,
shown here, also to an extent IL-12, IL-23 interferons, etc., etc., GMCSF. So it really does raise the question when you see efficacy in a patient, or if you see efficacy in Leo, which are the cytokines that you’re… what
is the basis of the efficacy that’s making Leo a happy dog vs. our initial views of it inhibiting strictly adaptive immunity?
So, in fact, the first generation Jakinibs really have broad effects
on innate and adaptive immunity, and it really makes you wonder of all these
different cytokines that are being targeted, really what’s the basis of efficacy? And then going forward, of course, you think, “OK, how can we improve that?” And
that brings the issue to the next generation Jakinibs. And so now there
is good evidence that drugs like Tofacitinib which target multiple JAKs and,
therefore, multiple cytokines have efficacy, and now that’s joined by other drugs
like Baricitinib that target JAK1 and JAK2, and again this drug also has efficacy as shown in phase 3
clinical trials in rheumatoid arthritis. But now added to that, drugs are being
developed in which you target a more restricted panel of JAKs. Filgotinib is a
drug made by Galapagos in Belgium. This is an AbbVie drug, I think, which targets JAK1. You can see that it targets a narrower spectrum of cytokines, and Decernotinib is a JAK3 selective inhibitor in which the phase 2A clinical trials were
published and now the phase 2B clinical trials are now online…or soon. So this is,
to me, this is a really exciting time in sort of in the history of Jakinibs
as to… you see all these drugs that have efficacy… How are they working? Where are they working? etc., etc. Now it’s a little, it’s a critical question again as
people are saying, well this is a JAK1 inhibitor, this is a JAK3 inhibitor.
Of course, like I said before, OK, I want… really… you to prove that to me in an in vitro assay with a recombinant
enzyme in a cell, and… but most of all… in people. So… and this is sort of the latest
scorecard of Jakinibs including the three FDA-approved inhibitors, Ruxolitinib, Tofacitinib, a canine JAK inhibitor, Baracinitib, in late-phase clinical trials, Filgonitnib and I actually don’t think I have the ABT compound here, but suffice it to say there’re many JAK inhibitors that
are being developed, and we may very well see in clinical… broad
clinical use soon. But I just wanted to take a few seconds and think about…OK,
where did we think things were in 1995? And how did it actually pan out? So
when we made our bold claim in the Science article in 1995, we weren’t
really… we predicted the efficacy; that seemed like an obvious thing to predict. We obviously had no idea of the safety, because, you know, the drugs were
inhibiting cytokines that are important for host defense. Now, at the time, if
anybody had asked me did I think selectivity was essential, I think
everyone in the field would have said selectivity would be critical, you know, assuming you could get it. So
that’s changed a lot in the last decade or so…couple of decades. Again, chemists can
seemingly easily do this. I mean, I couldn’t do it. But chemists in
pharmaceutical firms are very good at this. But I think, you know, no one would
have predicted that a Pan-JAK inhibitor would be viable, and so a drug like Baracitinib
which inhibits JAK1 and JAK2…maybe I’ll just go back for one second.. so Baracitinib inhibits JAK 1 JAK2 and really…
just recall that this family of cytokines 2, 4 7, 9, 15 and 21 are also JAK1
dependent, so if you had a JAK1 and JAK2 inhibitor, in principle, it should
do pretty much the same thing as a JAK3 JAK1 inhibitor, and you would be a bit
worried as I was saying that the knockout of JAK2 was embryonicly lethal…
knockout of JAK1 was perinatally lethal, so I don’t think anyone would
have predicted any, you know, a vice president at a pharmaceutical firm would be
particularly excited if their group came to him and said, “Well, we got an inhibitor that
pretty much inhibits all the JAKs, but don’t worry about it. It’ll
be a fine drug.” So, and in particular, I think people would have
been very concerned and companies were concerned when they had JAK2
inhibitors because of the lethality of the knockout mice, but I think the answer
now is that a drug doesn’t make you a knockout. A drug …they’re given doses. And so on one level the genetics really help you think about what a drug can do,
but, in fact, they don’t actually predict what’s gonna be useful in people or not.
I certainly predicted back in those early days that Jakinibs would be useful for
diseases mediated predominantly by adaptive immune mechanisms, but as I alluded to, I never really even thought about innate mediated disease. I mean, even though Charlie Janeway was telling me about this in this meeting in
Stockholm that I was telling you about. But, again, I wouldn’t have thought that it
would be such a great idea to be targeting a lot of cytokines. You would have
predicted back then that one of the side effects of JAK inhibitors would be infection… that we would
expect. If you had a JAK2 inhibitor you would expect that you’d have anemia and cytopenia, but what you wouldn’t predict was that necessarily was
a side effect that could be handled. One of the other side effects of JAK
inhibitors is hyperlipidemia. Now again lot of work on that… currently Tocilizumab is also associated with hyperlipidemia… it’s a very interesting area, but again,
something we didn’t predict. And again, going forward, I think we don’t know at
this point whether a JAK selective inhibitor will be better than the first
generation inhibitors or not. That obviously needs to be tested. So again, just…let me just run through some things that I think are lessons from this twenty
year odyssey of watching this pathway evolve. I think you know we still don’t
know the best uses for JAK inhibitors. I showed you a couple examples. I showed
you Leo and his dermatitis, and I showed you some other dermatological disorders.
So I think we’re still learning. I think, you know, I thought about this before
that from the time steroids were developed to the times where we really
thought we were good at knowing how to use steroids. When I came to the NIH in
the 1980s it was forty years or so after steroids were developed, and people were still thinking about strategies using alternate day steroids, pulse steroids, all this sort of thing. I hope it doesn’t take us 40 years
to figure out the right way to use JAK inhibitors. I would hope that we
can speed things up a little bit, but it does really give you some pause to
think that, you know, as we’ll get to, we always think we know more than we
actually know. The other lesson is that rare patients are informative on
multiple levels; knockouts are informative too, but they
don’t necessarily predict the risk benefit ratio for drugs in humans. And
this point I think I made several times, discovering new therapies and developing
new therapies takes a very long time, and then in the meantime science doesn’t
stand still, and our understanding of biology changes a lot, and I gave you a
few examples of that, and I think one thing that scientists often have… they’re
lacking… is humility. It’s not… we don’t… it’s not a big part of our
training to sort of promote humility, but actually, I think we really need
to embrace our ignorance and really fully understand that, you know, just ’cause you start here and know that the project’s gonna go here, that we’re going
to learn a ton during that time, and and that’s not to say you should be
paralyzed by that, but you should really have that… be thinking of that… and
what that does is provide opportunity for reconsidering mechanisms of action along the way, that even though we… our initial view is, “Oh well, this is how this
drug works in this disease,” you know, you’re gonna learn this enormous amount,
and it’s good, to keep going back. And what I would say to the
graduate students and trainees is, “This is great. This is job security, right?” I mean if we get all these things… if we don’t understand them …that’s good news for you… you’ll have jobs and be able to explain to us how we were wrong, and how
things really work. And that’s fine. OK, let me now turn to the second part of my talk
where I told you at the very beginning I was interested in this process: How does
an extracellular signal tell a cell what to do? And we, of course, all know from
the AIDS HIV epidemic of how important CD4 T cells are, but what we’ve
learned in the interim is that there’s all these external signals and CD4 T
cells are particularly good at sensing their environment to respond to
different types of infections, and we’ve learned a lot about so-called signal
dependent transcription factors through the discovery of STATs, and equally we’ve learned about lineage defining transcription factors also
known as master regulators and how that drives different types of lineage
commitment in CD4 T cells. A bit of a puzzle at the time was that even though
all of this looked very nice and very well-behaved, CD4 T cells, in fact, displayed a great deal of plasticity and have the capacity to
change lineage and all these sort of things, and, more importantly, also expressed more
than one master regulator or lineage defining transcription factor. In fact, there
is fairly flexible expression of master regulators. And co-expression of master
regulators so that the cells now, somehow, they were serving more than one master. So it really required us to rethink: What does lineage mean? How to
assess and how to think about transcription factors? So the good news
is technology came along and one of the things that we, of course, were able to
do in the early days with microarrays but now more recently with an RNAseq was to understand all of the transcriptome of cells that were differentiating. And
so there’s a lot of work on this. I’m going to tell you a very quick story on
this that gave a very unexpected result. And in particular we were
interested in this cytokine IL-27. It’s an anti-inflammatory cytokine; that’s its
requisite in vivo function. But it was really first described as a TH1 cytokine.
It’s knockout was lethal when you infected mice, and that’s because of this
critical anti-inflammatory function discovered by Chris Hunter. It suppresses
TH17 responses and it promotes IL-10. OK, so far so good. But here was a bit of a
mystery. It uses a receptor subunit GP 130 that’s also used by IL-6. So IL-6 is the prototypic pro-inflammatory cytokine, and it
induces TH17 cells. It promotes…it’s said to promote… TH2 differentiation, whereas IL-27 induces TH1 differentiation and inhibits TH17 differentiation. So we
thought this was a really interesting problem to sort of think about… OK, how
do we really understand specificity in cytokine signaling? So, again, if you’re
busy and have something to do, I’ll tell you … the paper got accepted and it’s published, has some interesting results, but the truth is we really don’t understand this.
That’s the short…you can take a short nap if you want, but I’ll go through the data
anyway. So, you know, what did we do? We did RNAseq to try to understand really what does this mean? These two cytokines that were so different, how are they acting?
What is the role of the… they both activate STAT1 and STAT3. How does that all work? So again, first… by first glance you’d say, “Oh, OK, you know, look, this is a heat map, RNAseq, look at all the genes that are
induced by IL-27, IL-6 in red, all the genes that are repressed by IL-27, repressed by IL-6.” And you’d think, “Perfect, they’re really exactly what we expected. They’re completely opposite.” Except what I’m not telling you is this: This is the whole transcriptome. All that’s different is this and this. So if
you look here you’d see, well, there are similarities between IL-27, IL-6. There are
differences. But where you really see these opposite effects, that’s only a
very very restricted part of the transcriptome. So that was a little
disturbing to us to see exactly the way you had shared induction of gene
expression, far more overlap than we originally anticipated. And then,
of course, now we have all these really cool tools like GSEA where we can, you
know, compare things very quantitatively and I said to Kyoshi who’s the postdoc doing this, “OK, OK. We’re just not smart enough to
look through those heatmaps… use some of these great tools and this will explain
everything to you. You’re gonna take IL-6 and IL-27 genes, and you’re gonna compare them to TH1 and TH2…
vs. IL-27.” So again, here’s TH1, TH2. And you can, Kyoshi showed me this, he said, “Oh, yeah. This is great. IL-27 gives you a TH1 response, and IL-6 gives you a TH2 response.” Then we said, “Oh, but what about this?” IL-6 also gives you this TH1 response. And IL-27 suppresses… easier
here… IL-6 induces a TH17 response, and IL-27 represses that. And if anything it
gives you sort of a T-reg reponse. But you see here IL-27 also induces a bit of a TH17
response, based… when you have all these genes. And so, at first, I said,
you know, what any good PI does to the postdoc when he comes back and says, “This doesn’t make any sense….” “Kyoshi, go back, do it again, this can’t be right.” But it’s right. Kyoshi went back and then we went through it all in our lab meeting. Kyoshi was right. And so, what I
interpret this to mean is that cytokines are sui generis. I mean, they
are what they are. I mean, if you say, you know, IL-27 is more TH1-like than IL-6…I’d say…barely… you know… not a lot. I mean… and again… just what I’m ultimately gonna say, in a way, is that maybe these small differences, if one of the small
differences here is interferon gamma, well, that’s huge. You say, “Is IL-27 more
TH1 than IL-6?” And you’d say, “Yeah, you betcha. It regulates interferon gamma.” But if you
look at it this way, you’d say, “Not so much.” And so I think both of those answers are
true, and it just, again, reflects our ability to think about this. So what are
the mechanistic underpinnings of this? Again, I’m running out of time. I’ll try to be
quick. If you look at the transcriptomes that are introduced in wild-type cells, and
then you look at cells that are… lack STAT3, you could see that the transcriptomes are entirely dependent upon STAT3, and that was a surprise to us.
We would have thought that IL-27… that STAT1 would have a more profound
effect on IL-27 signaling. In fact, when you remove STAT1, you have this very
strange effect that you now allow both cytokines to access STAT3 more
effectively, so you actually change the transcriptome by the loss of
STAT1. But, conversely, if you were fortunate again at the NIH to have this nice
ability to do bench to bedside… this is a disease called chronic mucocutaneous candidiasis, which is due to STAT1 mutation, and we have these patients thanks to Steve Holland at the NIH, and we
wanted to look at the effect of STAT1 gain of function on IL-6 and IL-27
transcriptomes, and that’s shown here. And you can see that, ordinarily, as I
alluded to before, that if you take cells and stimulate them with IL-6 and IL-27, they’re way more similar than we would have predicted. But what happens is in the
gain of function patients, you have more now diversities. STAT1… increased activity of STAT1 makes
the cytokine seem more different. And that’s quantitated here. So let me
just summarize that and say that there is way more overlap in the
transcriptome than we would have predicted, and this is very interesting ’cause my colleague, Bill Paul, who I’ll mention again at the end of the talk, predict… noticed this in 1989
and noticed this sort of oddity that when we were first thinking about cytokines we
expected them to have very specific functions and, in fact, they disturbingly
had this tremendous amount of overlap. And so, the TH-ness, we would say, of
cytokines can be really overstated, but it could very well be that it’s the
small differences that matter. And so when you have the whole transcriptome,
ironically, what goes… what’s important is to go back and understand exactly how
that cytokine regulates this particular gene. And as I’ll show you in a moment, the good news is now we can do that. The transcriptional output is driven by STAT3 whereas the ability to access STAT1 is important for specificity, and that’s
really revealed in these patients with STAT1 gain of function. And that’s
exciting because STAT1 and STAT3 levels are very variable but again
it’s not something that people thought about in the early days of the JAK-STAT pathway, but now people like Chris Byron think about this a lot and George Starr.
So again this is in cartoon form what I just told you here that STAT3 really
regulates the total transcriptomic output whereas STAT1
really regulates the specificity. OK, so the next step in this project sorta
happened in 2007 when, after the genome was completed, and when we got these machines, Alumina machines, to do next generation
sequencing. So now, here, you could return to this problem again and say, “OK, how
does the cytokine impact a cell transcriptome? That’s straightforward. But
where is a transcription factor like STAT3 going in the cell to mediate that?”
And the answer, in one sense… this is cells stimulated under TH17 conditions, looking at STAT3… and you could see STAT3 binding to the promoter of the IL-17 gene, STAT3 binding to the promoter the IL-17f gene and again STAT3
binding to the T cell-specific transcript of Ro-gamma-C, the
master regulator of TH7 cells. So, again, all of that was, to me, was just
tremendously exciting. I couldn’t believe it… that in my lifetime I’d be able to say, “Look
at the entire genome and find out where a STAT goes.” So that was exciting in and
of itself focusing on the genes, but as I’ll get to…
you also had this other stuff, like, oh well, you know, STAT3 is binding here. And some
people at the time said, “Chip-seq… you know, STATs bind everywhere in the genome… all these transcription factors. What they do? How important is
it?” I’ll come back to that in a second. So, we identified at that time thousands of
STAT target genes in TH17 cells and TH1 and TH2, etc. And the targets were all the things that
you were really hoping for… products of TH17 cells, and we were able to, in that case, find a direct link between the STAT, the
extracellular signal, and then what genes were being transcribed. But you
also saw this: Lots of binding in other places in the genome. And the other thing that you
could do… start to do at this time… was to do Chip-seq with things like histone
modifications. And what you see is that you see modifications of histones where
there’s accessibility of loci at the same places that STATs are binding in this
case. So that was the other big surprise when we got the genome, is that the genome is mostly not genes. And, you know, at the time, it was a little bit disturbing
that only two percent of the genome is genes, and then you’d say, “What is the rest of this stuff?” And at the time people said, you know, “This is just genomic junk.”
I would say that it’s still a bit of an argument. There are people that argue for the junk hypothesis. It turns out that we’re not the biggest genome. I
think onions are way bigger than our genome. It’s cool, right? ‘Cause onions have so many layers, right? You know, it’s like, they’re complex. We should have known that. But I think that lungfish, I think, I may be wrong… this is being
recorded so I should be careful… make sure what I’m saying is true… but I think
lungfish have the biggest genome of all. So that’s really disturbing that, you
know, that all of the complexity of us, if you want to ascribe it to genes, it’s
it’s a little disturbing, that it’s only two percent of the genome. But initiatives like
ENCODE have shown that 80 percent of the genome is active, including a lot of the genome
is transcribed in areas… places where there are not conventional
genes. And most of the genetic risk of disease is actually in the junk and
not in the genes themselves. So again we were very excited about this
and we were thinking about what does the junk tell you about, again, this basic
problem that I was thinking about, you know, way back when Marty Rodbell was
telling me about signal transduction. So the other thing that was discovered a
little bit later is this… that you can define what seems to be an enhancer by
this signature, and that’s histone 3 lysine 4 monomethylation is a mark
of poised enhancers, and then if you have histone 3 lysine 27 acetylation induced
by enzymes like p300, and you can actually have transcription of a different
type of RNA, an enhancer RNA, that’s the mark of an active enhancer. So we
became very intrigued by this. Golnaz Vahedi in my lab decided to enumerate this and came up with this very surprising result, or, at least surprising to me, which is that TH1 and TH2 cells which started out a few days before in vitro
had twenty thousand or so genomic switches, but only half were similar
between TH1 and TH2 cells. So ten thousand separate switches in TH1 and TH2 cells,
and basically no switches that were shared between TH1 TH2 cells macrophage and embryonic stem cells. Now, again, as luck would have it, that fall, in the same
issue of Cell, others were doing similar things and came up with pretty much the
same analysis that you had this tremendous diversity of switches and also that the major role of the environmental sensors, the, in this case
the STATs, were really the thing that we’re driving the enhancer landscape, and
the lineage defining transcription factors had less of an impact, so that, I
think that the prevailing wisdom prior to this was that if you’re a lineage
defining transcription factor, you should really supervise the epigenome when in fact that really is
not the case for some of the things that we call master regulators in T
cells. And Foxp3 is a good example of that, as Sasha Rudensky showed that the
Foxp3 comes in after the genome is modified. So the other thing that we’ve learned just very recently here is the distribution of enhancers in the genome
is very asymmetric and this is the work of Rick Young and my boss,
Francis Collins, and we had a paper earlier this year. And what you see
is if you look at binding a mediator K27 or p300, you see this very asymmetric distribution of enhancers. Now, to me, I grew up in New York, as you heard, this wasn’t so
surprising to me because I knew that Manhattan had these big skyscrapers, and then I went to college in upstate New York, and there were no skyscrapers.
So I get that… this is a little part of Manhattan. It’s dense, you know,
where a lot of activity is, so to me this seemed like it made a lot of sense. Now I should say I wrote a review and I used that metaphor, and I think they didn’t like it, but I like it.
You tell me whether you like it. Now again this region also here is enriched for GWAS hits, and the proposal has been made by Rick Young that these… the appearance of super-enhancers was linked to the idea of cell identity. And so we really, in Golnaz’s
paper here, we’re very interested in understanding the most relevant T cell
switches, because our thought was this is really telling you, you know, what are
the Manhattans for T cells or any given cell? Manhattan will be in a different
place, but, you know, what you’re looking for is you’re looking for the skyscrapers.
And so our idea was that this really was a nonbiased way of
identifying cytokine-dependent regulatory nodes. Again, this whole idea of
having its signal outside the cell… what is it telling the cell to do? How does it
do it? And where is that action? So, again, we repeated… we re-analyzed Rick
Young’s data in this paper, and we found just like Rick Young found, that in
embryonic stem cells the part of the genome that has super-enhancer architecture are transcriptional repressors, and that makes sense because embryonic stem cells are mostly about what they shouldn’t be doing. In macrophages and in T cells the
regions that have super-enhancer structure are chemokine, chemokine receptors and
cytokines and cytokine receptors. So on one level this was very exciting. On the other hand, you know, if you’re an
immunologist, you sorta say, “Really? What are you expecting? Of course T cells are regulating cytokines. It’s a very expensive way to figure out what you
already knew.” The paper’s published. I’m not going to go into a lot of the detail of
this. We did have a big surprise which was the super super-enhancer is
a gene called BACH2 which, fortunately, we had just written a paper in Nature a
couple years ago on how BACH2 is this critical regulator of tolerance in T
cells, and in all subsets of T cells the thing that scores the highest super-enhancer structure is BACH2. So there were some obvious things and then some less
obvious things. And the other thing that’s less obvious is that long
non-coding RNAs are enriched for the super-enhancer architecture. But this part…
at least that part makes sense. I like things when they make sense too. So the other thing that we found that was sort of a reiteration of Rick Young’s data is that if you look for loci
that were associated with autoimmune disease, they’re enriched for
super-enhancer architecture in T cells, and so you can see examples… we have this enrichment of super-enhancer architecture and the nice comparison is
type 1 and type 2 diabetes; here you have an enrichment of T cell super-enhancer,
and here in type 2 diabetes, a disease you usually don’t think of as having a heavy
immune component, you don’t have this tremendous enrichment. So the other thing that we were excited to
see was that if you treated cells with Tofacitinib, you could see that Tofacitinib is primarily acting on loci that have super-enhancer architecture
compared to genes that have typical enhancer architecture in these violin
plots. You see these very dynamic changes here. OK, to summarize this part, you have dynamic tissue selective expression and
that correlates with super-enhancer architecture. Cytokine and cytokine receptors are the dominant class, but BACH2, what we refer to as a guardian transcription factor, is really the champion. And so it really has… it really
does have this ability to supervise the genome and prevent aberrant transit into
effector cells. Polymorphisms associated with immune-mediated disease are more
likely to have super-enhancer architecture, and typical enhancer architecture and
rheumatoid arthritis-associated genes with super-enhancer architecture are
preferentially affected by JAK inhibitors. So I’m running out of time,
but I just wanted to summarize a few things. So membrane-to-nucleus signaling has gotten very interesting since the 1980s. I mean, Marty Rodbell’s discoveries in this
class is amazingly exciting just linking the biochemistry, but we’ve made
enormous advances in taking the first steps in the plasma membrane to getting it all the way to the nucleus. And we’ve learned a lot about the elements,
so this is a very exciting basic advance, but as I showed you it also
provided lots of opportunities for new therapies. So I think, you know, as I was talking to students earlier, I was thinking the challenge, of course, ahead, in a way,
is now we can measure everything, right? We can measure transcriptomes; we can
measure epigenomes; we can measure changes of proteome, etc. And so I
now think that the real challenge is integration of signals that impact chromatin organization in real time. So we should be able to do this. Most of the things I told you about were snapshots. So, you know, increasingly, we will be able to do these things on single cells; we will be able to do them over time. And really,
this is what I would like to understand… is how do extrinsic and intrinsic
signals really modify the chromatin architecture to enable
transcription in things I don’t have time to tell you. We’re now beginning this with techniques like ChIA-PET, etc., etc. But the
exciting thing is we have great new tools like CRSPR/Cas(9) and super-resolution
microscopy. I think we’re really at an exciting new sort of place for really
now understanding signaling in a way that we couldn’t possibly dream of. And it
should be a lot of fun to do this. I’m really looking forward to the next
twenty-some-odd years of studying signaling. I’ll stop right there. I did want
to acknowledge Bill Paul. Bill Paul has been a tremendous mentor to me over the
years. Sadly, if you know, Bill died earlier this month and was just this
terrific person who really really pushed immunology both at the NIH and around
the world. And I wanted to sort of honor him. Lots of other people to thank along the way… lots of people in my lab I think I mentioned them. Also collaboration with Francis Collins on super-enhancers I mentioned. But, again, NIH, an amazing place where people are
extraordinary collaborators and willing to share their expertise. And
then a long very satisfying collaboration with Pfizer, especially Paul Changelian
and now more recently with Jim Clark and J.B. Telliez. So thanks very much. I think I’ve run out of time. Thanks.

Leave a Reply

Your email address will not be published. Required fields are marked *