Imaging Estrogen Receptors and Estrogen-Receptor Function in Breast Cancer by Positron Emission Tomography (PET): Improved Selection of Patients for Benefit from Endocrine Therapies
Authors: John A. Katzenellenbogen*, Michael J. Welch, Farrokh Dehdashti, and Barry Siegel
The growth of many breast cancers is regulated by estrogens, and these cancers can be treated effectively with endocrine therapies (ET). Identifying patients who will benefit from ET, however, continues to be challenging. Measurement of estrogen receptor (ER) levels in tumor biopsy samples using immunohistochemistry provides accurate negative prediction (no ER implies no ET response), but poor positive prediction. We have developed a fluorine-18 labeled estrogen, 16alpha-[F-18]fluoro-17beta-estradiol (FES) for in vivo imaging of ER in breast cancer by positron emission tomography (PET). Clinical PET imaging studies with FES provides a higher positive predictive value for benefit from ET than does IHC. In addition, an endocrine challenge test that measures the change in tumor metabolism, assessed by PET with 2-[F-18]fluoro-2-deoxyglucose (FDG) 1 week after tamoxifen or after 1 day of estradiol therapy, provides a highly accurate prediction of the ultimate benefit of ET in treatment. Both FES-PET and serial FDG-PET are promising approaches whereby it should be possible to tailor ET to individual breast cancer patients.
Thank you very much.
I really appreciate the opportunity to talk to you this morning about what you see here.
This is actually a long-term collaboration I've had with people at Washington University
in St. Louis using positron emission tomography to image estrogen receptor levels and estrogen
receptor function with the specific purpose of better selecting patients who will benefit
from endocrine therapies.
Now I think you all know that estrogens are female reproductive hormones.
They also have effects in men and they affect many tissue systems and organ systems outside
of the reproductive tissues.
But the mammary gland is the one we'll focus on today because estrogen promotes mammary
gland development and also drives a certain fraction of breast cancers.
Now the estrogen receptor is a transcription factor, ligand binds to it, and it binds
to specific DNA sites and alters patterns of gene regulation.
So as a target for imaging, it is a intracellular, actually intranuclear target.
So anything that would detect it by labeling a small molecule would have to penetrate the
cell.
Now in terms of the role estrogens play in breast cancer, I mentioned that it's a driver
of proliferation.
And this was actually recognized indirectly more than a century ago that 30% of breast
cancer patients would benefit from some form of endocrine therapy.
At the time, it was actually removing a woman's ovaries and a very alert Scottish physician
recognized this.
And this really can be considered the first form of targeted therapy way back then.
Now that ovaryectomy has been replaced by endocrine therapies that are anti-estrogens
or aromatase inhibitors.
And if it works, it spares the cancer patient the morbidity associated with chemotherapy
or radiation therapy.
But only a third of patients respond.
So the question is, how do you identify those patients?
And when the estrogen receptor was first characterized in the late 60s, they figured that was the
answer.
If you have receptor, you'll respond to endocrine therapies.
If you don't, you won't respond.
Now it turned out that half of those statements are true.
Because if you look at estrogen receptor levels by immunohistochemistry assays, which is what's
done now typically, only a third of them lack receptor.
And this negative test is actually very predictive of a lack of response.
Two thirds have receptor, but only half of them respond, and so the other half don't.
So the estrogen receptor assay, as typically done, has strong negative predictive value,
but poor positive predictive value.
And we speculated that if we could image estrogen receptor, and particularly estrogen receptor
function, we might have a better predictive value.
Because you really want to know whether the receptor is functional.
One of the functions is binding hormone.
That can be assessed in vivo by PET, but was initially assessed on tumor biopsy samples
by biochemical assays.
And now in a way has sort of regressed to look for just some sort of protein antigen.
Now we worked to get ways of labeling steroids with isotopes that can be used for positron
emission tomography.
Fluorine 18 is a very convenient one, has good nuclear characteristics, has a two hour
half life.
And this compound called FES has a fluorine at the 16 alpha position.
I'll come back to this other one in just a minute.
Now in a sort of a qualifying, well this is how it's made, obviously the synthesis has
to be rapid because of the short half life.
This was our initial synthesis where we displaced the triflate and did a reduction to get FES.
There's another synthesis that was developed about a dozen years later that's actually
somewhat more convenient.
Now just to test whether this compound actually was selectively taken up by target organs.
If one uses an immature rat, which has low endogenous estrogen levels, FES is taken up
very avidly by the uterus and ovaries, but not by non-target tissues.
And it's blocked very specifically by an excess of unlabeled estradiol in these tissues, whereas
the non-target tissues are not blocked.
Now the liver and kidney have high levels because most of the dose is fluxing through
those organs as part of roots of excretion.
So it shows the characteristics of selective uptake in a target and selective displacement
from the target.
The question is, will it really realistically image ER positive breast tumors by PET?
And PET, if you don't know what it stands for, it stands for positron emission tomography.
Certain isotopes that have excesses of protons will decay by a proton converting to a neutron
which ejects an antimatter particle equivalent to a positive electron.
It annihilates with a normal electron, generating two energy photons that come out precisely
in opposite directions at the same time.
And if you have a device, a positron emission tomograph, which is a series of detectors
surrounding a patient, you can look for coincident events on detectors opposite sides of this
circle and you can reconstruct an image in three dimensions.
And this was an image we obtained some years ago with fluoroestradiol.
And you can see a primary tumor.
This is a transverse image, basically what would be a hypothetical slice perpendicular
to the axis of the body.
Primary tumor, chest wall metastasis, and the liver because, again, a lot is fluxing
through the liver.
Now, we heard yesterday a very nice talk by Dan Heller about how you can image things.
And imaging tumors is really no great fate.
I've sometimes said if you label dirt, you could actually image a tumor because it picks
up particulates and lots of things.
So is this a meaningful accumulation that is dictated by the presence of receptor?
So we did find that if we took out tumors and measured estrogen receptor concentrations,
that the intensity of the image did correlate with receptor concentration.
But the real question is, does it predict better the response to endocrine therapies?
And here I return to the slide I showed at the start.
So this is the target that we have, the receptor, and we have fluoroestradiol to image this.
But twice as many patients have receptor than respond.
So the question is, was there some way we could also image response?
And so this is what's typically done now.
If a woman has an abnormal mammogram and biopsy is obtained, receptor levels are measured
by immunohistochemistry.
If they're receptor positive, they're often put on the drug that would block receptor
action in some way.
And the clinical outcome is assessed by radiological methods or standard clinical criteria.
This can take several months to establish whether the therapy is effective.
And if it isn't, one's lost time and proceeds then probably with somewhat less good outcome.
Now what we're proposing to do is to query the receptor in the tumor itself with fluoroestradiol.
And also to look at a cellular response, the uptake of fluorodeoxyglucose, which is a general
tumor imaging agent and measures glucose demand.
And those tumors have hyperglycolytic activity, so this is a general tumor marker.
Now you might expect if the therapy is effective that this would go down with therapy.
But we were looking for something else, which I'll mention in a minute.
So what was done is that advanced breast cancer patients were first imaged for receptor and
cellular response or metabolism before any therapy.
They were started on tamoxifen, which is an anti-estrogen that still has a little bit
of partial agonist activity.
They were imaged then after a week or 10 days for receptor, and we would expect tamoxifen
would block receptor, also for glucose.
They were continued and assessed for clinical outcome by standard therapies, standard assessments.
Here we look for a decrease to indicate, like in the parallel with the rat uterus, that
we could block the uptake if it was due to receptor.
Here we were actually looking in responders for an increase because it's known clinically
that some patients who are feeling pain will feel worse for about a week on tamoxifen and
then feel better.
This has to do with a transient stimulation of the tumor as tamoxifen levels are building
up because it's somewhat of a partial agonist, particularly at low concentrations.
We were looking for a metabolic correlate of clinical flare.
Clinical flare is not very useful because if you don't have symptoms, you don't sense
it, or the disease may progress so you may feel worse but never be responded.
Here are some images before tamoxifen.
This is a receptor image.
This is a glucose image.
This is standardized uptake value, which is a ratio with what would be uniform distribution.
This would be five times greater than that.
After tamoxifen, the receptor image is blocked as we had hoped.
In this patient, we saw actually an increase in the intensity of the glucose uptake, which
we would classify as a metabolic flare.
Another patient, a really clear indication of three tumors with the receptor agent rather
equivocal image based on metabolism, a complete blockade, 90% blockade after tamoxifen, and
here an increase.
Now, 40 patients went through this, were classified as responders and non-responders by oncologists
who were blinded to the results of imaging.
This is how the responders and non-responders turned out in terms of the fluoroestradiol
uptake.
If we say SUV of two is a cutoff, we've now improved from no predictive value.
These are all ER positive patients, a positive predictive value of nearly 80% and a negative
predictive value of nearly 90%.
The in vivo assessment of the capacity of the tumor to take up an estrogen binding activity
is better than the immunohistochemistry assay.
What was particularly striking is that the difference between the initial and the final
glucose measurement a week later showed that if you had a 10% or greater increase, you
had really a 90% chance of being a responder, whereas if you were below that, you had a
95, 94% chance of being a non-responder.
So we've really greatly improved the predictive value using PET by looking at the presence
of the target, whether the estrogen receptor is capable of binding ligand, but also particularly
querying whether the receptor is able to mediate this transient stimulation seen with tamoxifen.
This is a hormone challenge paradigm that measures function that I think can really
be generalized and really increase the power of prognostic and predictive biomarker imaging.
Now, we've done an additional study.
So that took a week, two images in a week, to tell whether a patient would respond.
We did another study in more heavily pretreated patients who are now going on other endocrine
therapies that would not cause a flare, more complete anti-estrogens or aromatase inhibitors.
And we did a different sort of challenge.
So we imaged glucose and estrogen receptor levels as before.
These patients were then given one dose of estradiol for one day and imaged again for
glucose.
And here you can see an enormous increase in glucose uptake just after one day in what's
a responder.
We didn't reimage the receptor because it turned out that the second image after an
endocrine challenge was completely blocked and didn't add value.
And it turned out in this study with the heavily pretreated patients, they were all considered
ER positive based either on history or more recent assays by immunohistochemistry methods.
Only a third was expected to respond, and we got a third responders.
Now here the receptor image intensity using the old cutoffs was less predictive than before.
Not surprising, these were more heavily pretreated patients.
But the functional endocrine challenge, now just one day of estradiol, really separated
the responders from the non-responders very effectively.
And so again, this is a measurement of the function of the receptor.
Even in these advanced patients, we could tell in one day whether the system was functional
and was likely to respond to these more advanced endocrine therapies.
And so the change in glucose uptake is predictive even in these heavily pretreated patients.
It's much more predictive than the ER immunohistochemistry assay as well as assays of other targets.
Now I think the conclusion to make is that the test of function, if it's possible, will
be more informative than the test of presence.
If you have targeted therapy, you always need the target, but the target has to be functional.
And if you look at a lot of assays, a lot of methods like Herceptin for HER2, the target
has to be there.
But even if the target is there, you get a 30% response rate.
So if you can develop challenges to see whether the target is functional, that can be more
informative.
Now going forward, the last set of data I've plotted here as accumulation of non-responders
and responders as a function of change in glucose.
And they're really pretty well separated.
That's why we got those high predictive values.
But if you look at the responders and non-responders and imagine that you might have a 10% error
in your measurement, that includes about 25% of the cases.
And really you'd like to have a test that would separate the responders and non-responders
really quite dramatically.
So we're looking at other things that we could image that might be more robust predictors
of the functional status of the estrogen receptor.
And one target is the progesterone receptor, which is very acutely regulated by estrogen
stimulation.
This is a xenogast system, MCF7 cells grown in a thymic mice.
And you can see when they're withdrawn from estrogen and then re-stimulated, the levels
of PR, RNA, and protein go up very dramatically, as do other proteins.
But PR is something that we can image.
And I showed you this molecule, somewhat more decorated progestin, that can be used to actually
image progesterone receptor positive tumors in humans.
Now this is an early study that was just published where we can see that we can identify
PR positive patients, but the real study would be comparing before and after an estrogen
challenge to see if we get a more robust response that's even more predictive of the success
of endocrine therapies.
So a test of function, as I said, will be more informative than a test of presence,
if you can do it.
And really I urge all of you who are going into the cancer area, particularly the cancer
detection or imaging area, to focus on the medically relevant question.
Because there's a lot of imaging you can do that's sort of neat and neat technology, but
if you want to have an impact on medicine, you really have to understand what the medical
questions are and what the current limits on diagnostic methodologies are.
So I've had a lot of people over the years who've worked on this.
These are past co-workers, current people in my lab doing various aspects of PET imaging.
I mentioned at the start that colleagues at Washington University Medical School, Mike
Welch, and many of his colleagues have been integral in these studies, and these are the
funding agencies.
So thank you very much again for the chance to talk.
Any other questions?
It seems like a natural step to go from these molecules as imaging probes to therapeutics.
So, for example, linking to a vitreo molecule or...
Right.
Because the target is intracellular, you have to be a little careful not to perturb the
structure so much that you change the distribution properties.
We're looking actually at other halogens, so bromine 77 has very effective Auger cascades,
and because this is localized on the chromatin, sometimes you can get very effective local
damage by the Auger cascade, the high linear energy transfer of those low energy electrons.
So that's something that we're working on.
There are other people that are using iodine-123, which is a similar isotope.
But maintaining the very favorable target to non-target distribution and retention of
the target tissue consistent with the half-life of the isotope is also important.
So we have to optimize other characteristics.
We have time for one more question.
I was wondering if this was as predictive for a woman with stage four breast cancer
as with someone with DCIS?
At this point, we've not used it with DCIS, which would actually be really great if we
could do that.
I think that the actual mass of tumor tissue in DCIS is relatively small, though the levels
of receptor could be high.
We talked about how to do this, but at this point we haven't done it.
But it's really been the later stage cancers where we've gotten good predictive values.
Thank you.