Tip:
Highlight text to annotate it
X
Thomas Giordano: Thank you very much. I'd like to thank the
Planning Committee for the opportunity to present, on behalf of the Thyroid Group, our
work today. So, today we're going to talk about the view of the data from 30,000 feet,
but before I do that I felt obligated to give a little background.
So I was thrilled when the TCGA announced that they were going to do thyroid. Thyroid
cancer can sort of get shortchanged in the big scheme of cancer. I think one of the reasons
that they did is because thyroid cancer is clearly on the rise. However, the mortality
of thyroid cancer is not that high. It's actually -- over the same period, has actually trended
down flat and now is rising up. But, clearly, it's one of the few cancer types that is actually
increasing in incidence.
So, I'm a professor of pathology so I have to show some histology -- no, just kidding.
This is important because the papillary carcinoma, which is what our project's on, has three
main types, and it'll come through in the data so I really wanted to spend just a minute
to express this. So, papillary carcinoma gets its name from the tumor on the left here,
which is a tumor that has finger-like projections, or papillae. But not all papillary carcinomas
are actually papillary in terms of their architecture. There's a group here that we recognize that
are called the follicular variant, not to be confused with follicular carcinoma, but
the follicular variant of papillary carcinoma. And then there's a third main type, the tall-cell
variant, in which the cells are more columnar than cuboidal. And so we recognize these three
variants, and it's kind of important. We know from lots of papers that these have different
genetic profiles and different gene expression profiles, and that will also come out in our
dataset.
These are the common genetic defects in thyroid cancer. Now, we're just looking at papillary
carcinoma. This whole chart shows the whole spectrum of thyroid cancers, but it's worth
looking at. This one on here is post-Chernobyl, so radiation-induced cancers. So, RET rearrangements
are common in papillary carcinoma, even more common in radiation-induced areas. BRAF mutation,
a very high rate of V600E mutations; a small rate of BRAF rearrangements, which we'll see.
NTRK rearrangements. RAS mutations are quite interesting. So, notice it goes from zero
to 21 percent, so depending on the cohort of papillary carcinoma you'll get different
numbers, but we definitely know that the follicular variant is enriched for RAS mutations, just
as follicular carcinoma. PAX8 -PPAR gamma root rearrangements are thought to be present
mostly in follicular carcinoma, but there's been a few recent studies that show that they
can also occur in some papillary carcinomas; again, the follicular variant. And then β-catenin
and TP53 mutations are sort of things that occur down here when tumors show histologic
progression to poorly-differentiated and undifferentiated carcinoma. So that's the genetic landscape.
There's a big issue for -- a big opportunity, a big issue for the TCGA. So, if you genotype
thyroid cancers, differentiated thyroid cancers, you only find one of these driver mutations
in about three-fourths of the cases. So we don't know what that other mutation -- what
those other mutations are. So, this project is all teed up to try to find those other
mutations. And this is important, not just for understanding of cancer biology, but there
are already cancer diagnostic tests on the market based on genotype. So if we can expand
the genotypic universe of papillary thyroid cancer, those tests will get better.
So, again, this is the first look at the data. We did a data freeze about -- just less than
a month ago. Much of what I'll show you today was generated by the Firehose platform. We're
really just getting started, and Gaddy wanted me to stress that much of this is not been
validated.
Here's the sample counts. We're pretty far along in the project, and the remaining cases
are sort of in the pipeline. So we're thinking we might just wait until we full it out to
the full 500, but we're still discussing that.
Here's some of the clinical data. Thyroid cancer occurs in a younger population. The
mean age is 46. You know, I was sort of envious when I saw those survival plots. That's really
not so easy in thyroid cancer; there's not a lot of deaths. To do a good outcome study
in thyroid cancer you sort of need 15 years of follow-up, and so we're not going to have
too many Kaplan-Meier plots anytime soon. You can see we've had one death so far. Like
in all endocrine diseases, way more common in females, almost three to one. And we have
a good breakdown of the different histologic types.
So here's some of the data. So, again, I want to stress that papillary carcinoma is a differentiated
tumor, and overall it has a low mutation rate. So here's the mutations per sample. It's generally
uniform across here, but you can see there's a few samples that are jumping up and showing
that they have a much higher mutation rate. The sequencing coverage is outstanding. And
then down here we can see the rate is quite low. I'm pleased someone showed the plot earlier.
I don't know if anyone looked. I looked for thyroid. It was definitely on the left end
of that plot, away from the head and neck squamous cancers. So it's a low mutation rate;
not surprising because it is such a well-differentiated tumor.
So this is a great slide that integrates a lot of the data, and I'll spend a little time
going through this. This shows the common mutations, with BRAF here representing 57
percent of our cohort. We have some RAS genes, and notice the BRAF and the RAS genes and
the fusions down here are all mutually exclusive, which we've known in thyroid cancer, the thinking
being that, you know, having more than one of these doesn't add any biological advantage.
The histologic data is up here, and so it's hard to see, so I'll explain it. So the BRAF
tumors are enriched for the classical type and the tall-cell type, which is consistent
with the literature. The RAS mutant tumors are almost all the follicular variants. Again,
this is consistent with what we've known. The tumors with the fusions are mixed between
the classical type and follicular type. And then these so-called wild-type tumors, where
we don't know the driver mutation, have really a smattering of histologic types. So immediately
we're seeing a very strong correlation between genotype and histologic type. And so that
gives me a lot of confidence that this is a quality dataset.
We can see the fusions, as I mentioned briefly, are here. RET fusions are the most common,
but we're also picking up some PAX8-PPAR gamma, NTRK, and a few BRAF fusions down here. And,
again, mutually exclusive with BRAF RAS. I want to bring your attention to this point
right here, which is this initiation factor, and it's also mutually exclusive with these
other common mutations and the fusions. So that's telling me -- suggesting to us that
this is a biologically significant mutation even though we have not validated that.
And then the copy number changes are shown here. Notice there's -- across the BRAF cohort
there's not a lot of copy number changes. Across the follicular variant or the NRAS
cohort, there's not a lot, but notice there's a band right here that represents chromosome
22. So there's an enrichment for a loss of chromosome 22 in these RAS mutant tumors.
The tumors with fusions are also pretty quiet. And then the remaining 20 percent, the so-called
wild-type tumors, can be divided into two groups: those with more changes and those
that are pretty silent.
So the first thing that I want to do as a pathologist is I want to go back and look
at these tumors over here that don't have a lot of changes and just review their pathology
because, truth is, endocrine pathologists, we fight over diagnoses a lot, so we need
to look at that. There's also messenger RNA differences that we know, some genes here,
and then microRNA, which I'll show a little bit more.
So this tells a very compelling story, this one slide, that integrates much of the data
and shows that this is a quality dataset. Yes, we've replicated some things that we've
known, but the integration of all the data in this setting I don't think anybody has
come close to doing.
So this is that new potentially novel mutation, this X-linked translation initiation factor.
It's very interesting. There's no known role in thyroid cancer, and, in fact, we could
find just one other synonymous mutation in the COSMIC data base. So this will take validation,
but it's certainly an interesting finding.
Fusions -- I showed some fusion data. Truth is, we're still going through this, there's
much work to be done on fusions. This was some data from an earlier analysis that showed
this fusion right here, which is, in the business, called RET PTC1, which is the most common
version. And then we have some more novel things. This ETV6, NTRK3, which, in speaking
with Jim Fagin, he's validated this in an independent cohort from the Ukraine, both
in radiation and non-radiation cases, so we think this is a real finding. And then PAX8-PPAR
gamma as we'd expect in a few cases.
Onto the methylation work. The methylation profiling identifies four classes of tumors
that generally correlate, again, with histologic type and mutational status. So we can see
there's two groups over here on the left that are mostly the classical type and the tall-cell
type, and they are enriched with BRAF mutations. And on this side, which is more similar to
the normals, we have an abundance of the follicular variants and tumors with RAS mutations. So
the methylation work is integrating nicely into the same compelling story.
A few interesting molecules, miR-21, which we heard a little bit about this morning,
and miR-146b, both been worked on in thyroid cancer but not at the methylation level. And
here we show inverse expression between methylation and expression, so these are interesting leads
for us to work on.
And then, as a thyroidologist, we always like to look at thyroid-specific genes. Not because
they tell us that much about the cancer biology, but more they give us thoughts about progression
and possibly a loss of radioactive iodine treatment. And so that's a big issue in our
field when tumors become a little less differentiated and no longer respond to radioiodine. So,
here we can show one gene, thyroid peroxidase, that actually shows a differential methylation
profile based on the different types, classical follicular and tall-cell. And where from my
group and others, we know that TPL goes down in BRAF mutant tumors compared to the follicular
variant, and, sure enough, methylation is playing a role here, again inversely expressed
between methylation and expression. So that's a potentially interesting story, and we need
to look at other genes in the thyroid, like the sodium symporters, sodium iodine symporter,
and others.
MicroRNAs; we can use various tools to cluster these. Here's a cluster driven by miR-21,
so we can get four types or seven types. And notice the four types is actually showing
some, again, correlation to histologic type, which we know reflects genotype. So I think
the microRNAs will fit in nicely. This needs some work though. If you look at the British
Columbia software versus Firehose, you can generate different clustering algorithms and
then actually compare them head-to-head and start to make some sense of how much you believe
them. This is kind of interesting. So, this is the British Columbia software versus Firehose,
four groups versus three, and you can see some cohorts line up nicely here and other
cohorts get sort of scattered into others. So, we have some work to do to try to figure
out what's the most meaningful way to look at the microRNA data in a global sense. But,
clearly, there's some useful data in there.
And then cancer regulome. Lisa Iype was kind enough to prepare some slides for us in which
they look at all the factors, all the measurements throughout the dataset, and we just started
with something simple, like histologic type, and she was able to show that there's many,
many, many associations shown here. And, sure enough, some of the more interesting ones
are -- is miR-21, which I just showed you had differential methylation expression, and
here's BRAF. So these are all potential leads and we'll certainly be using some of these
software tools which are at our disposal.
So, in conclusion, I would argue that we're making good progress. I think that we're progressing
as planned. I think the cohort is outstanding and truly representative of the disease. I
make this point because that's not true for every paper in the literature. People cherry-pick
cases, and what they're publishing is not representative of the broader disease. So
I really do think this cohort is accomplishing that.
We have a low overall mutation rate, with few copy number changes, but we have a few
tumors that have increased copy number changes, which we'll spend more time working on. And
we've uncovered and reproduced strong associations between the tumor morphology, its genotype,
its gene expression profile, copy number changes, and methylation status, and we've uncovered
many interesting, novel leads, mutations, and gene expression patterns to keep us busy
for quite some time.
So there's much to do. We still have to decide whether we're going to fill out the whole
cohort or publish where we're at now, but I do think we're on track for our first paper
in the middle of next year.
As with all the projects, there's many, many people to thank and I would hate to go through
these individually, because I'm sure I've left people -- some people out, but, clearly,
I'd like to thank Gaddy Getz, who I've not met yet. So Gaddy, if you're here, come up
and introduce yourself. But thank you very much.
[applause]
Raju Kucherlapati: Thank you very much. You think that the fact
that so many of them have BRAF mutations would suggest that the BRAF inhibitors might be
a therapeutic approach for them?
Thomas Giordano: Well, you know, I wish -- you know, maybe
Jim Fagin can answer that. He's more in tune to the clinical trials. There's certainly
many clinical trials ongoing. Steve Sherman, at MD Anderson, is working on that. But I've
not heard of -- it's not distilled down into, like, this home run where people are ready
to give up radioactive iodine. Matt?
Matthew Meyerson: Sure. So just a couple questions and then
a comment. My first question is about the relationship between, you know, thyroid carcinoma
and lung adenocarcinoma, because I think there are a number of similarities. First, the major
-- you know, as you very well know, the major marker and one of the leading amplified genes
in lung adenocarcinoma is the thyroid transcription factor-1 gene, you know, suggesting potentially
some common etiology. And, you know, there's also some --
Thomas Giordano: And there's also RET rearrangements, too,
that are common.
Matthew Meyerson: There's RET rearrangements, BRAF, and RAS
mutations, and I'm wondering if you've started to look at commonalities between the diseases
or have thoughts about how to do so?
Thomas Giordano: You know, certainly that would be a good comparison,
but there's some jump -- things that jump out, like there's very few EGFR mutations
in thyroid. So I think the similarity is there, but I don't know how strongly it'll hold up.
Matthew Meyerson: And I think my second thing is more of a comment,
which is I think if the rest of the rest of the cancer community could do as well as the
thyroid cancer community has done, we probably wouldn't need TCGA or a National Cancer Institute,
so we're very struck by the survival data.
Thomas Giordano: Yes, yes. Well, that's not to belittle thyroid
cancer. Most of the deaths, though, are actually in those poorly differentiated and anaplastics.
If I showed those survival plots, the survival plot for anaplastic is measured in six months.
So, thyroid cancer represents the whole spectrum, and I'm hoping that we're going to get to
a more aggressive thyroid cancer project, which will then nicely integrate right into
this, because we have many cases that have both papillary carcinoma sitting right next
to anaplastic carcinoma. It'd be nice to profile both of those in parallel.
Thank you, Matt. I don't know if there's --
Stephen Chanock: Can I ask one last question?
Thomas Giordano: Yep.
Stephen Chanock: Thyroid cancer, you know, is one of the most
interesting ones with respect to heritability and familial syndromes and, you know, sibling
risk. And, you know, it's been in the past discussed that the outcome may be related
to whether the early-onset, family-driven cases versus the more sporadic later ones.
When you look at the dataset you've put together, are you able to, in any way, parse out, or
do you have information on the heritability, in terms of family-related cases or the like?
Thomas Giordano: So we have to separate, obviously, medullary
thyroid carcinoma --
Stephen Chanock: Right.
Thomas Giordano: -- which has a much stronger familial association
than, you know, sort of follicular-cell thyroid cancers. I know the group at Ohio State has
worked on this. There's some germline mutations of, I believe, TTF-1 that, you know, make
you susceptible to papillary. So there are familial cases of familial papillary carcinoma,
but it's not nearly as strong as medullary. That's something that I think we'll get around
to looking at, but it's not something that's going to jump out at us right away.
Raju Kucherlapati: Thank you.
[applause]