John F. McDonald, Ph.D. Professor, School of Biology, Georgia Tech CSO Ovarian Cancer Institute, Georgia Tech
Friday afternoon session: Bioinformatics and Systems Approaches to Cancer Treatment and Diagnostics, Beckman Auditorium, 3:25 PM -- 4:25 PM
Abstract: The field of cancer biology is rapidly moving from an era of specialization to an era of integration. In my presentation, I will provide examples of the integration of molecular biology with analytical chemistry, computer science/machine learning and material science/nanotechnology to develop novel approaches to the diagnosis and treatment of ovarian cancer.
Speaker bio: In addition to his responsibilities as Associate Dean at Georgia Tech, Dr. McDonald also serves as Director of Tech's Integrated Cancer Research Center (ICRC). ICRC researchers are using the latest genomic- and nano-technologies to develop new and effective cancer diagnostics and therapeutics. Dr McDonald is the Chief Scientific Officer at the Ovarian Cancer Institute (Atlanta) and previously served as Head of the Genetics Department at the University of Georgia (Athens) and Chair of the School of Biology at Georgia Tech. Dr. McDonald is an Elected Fellow of American Association for the Advancement of Science (AAAS). He holds a PhD in Genetics from University of California, Davis, CA. and did his post-doctoral training at the University of California San Diego (La Jolla) and at the Institute for Cancer Research at Columbia University Medical School (New York).
First of all, I want to thank the organizers for giving me the opportunity to participate
in this great meeting and to congratulate you on a fantastic job.
It's really impressive to see how the students, how good the students are here and how well
you are at organizing.
I'm going to go back with a mission to really kick our students' butts because they're way
below your level of proficiency.
Okay, so I'm going to start off by just making a philosophical statement or an observation
of the history of science that I think that at this particular moment in history, all
areas of science and technology are moving from an era of specialization into an era
of integration.
When I was a graduate student, you know, we knew there was engineers on the other side
of campus.
We didn't ever talk to them.
They seemed weird.
We knew they were doing something important.
We could tell who they were because they always had white socks and black shoes, except in
the summer where they switched to sandals, but they still had the white socks on.
We knew they were doing important things.
We knew the computational scientists were doing important things, but we were in our
own silo, in our own world, and we really didn't communicate with them very much.
And I think that it was important for going through that era of science because I think
we learned a lot by studying intensively in our own fields, in genetics, my field.
Of course, we learned a lot about molecular biology and all of that.
So we know all of that now.
I think the next era of discovery is really going to involve the integration of this specialized
knowledge.
I think areas and disciplines that traditionally have not communicated with one another are
already communicating, and that's obvious.
I think on this campus, and especially as manifest in this meeting, and I think that's
where the huge and big breakthroughs are going to come in the future.
So I think all of you students are in the right place at the right time, and I envy
you for starting your career at this time.
I'm on the way out, but you're on the way into a very exciting time in science.
All right, so given that sort of philosophical point of view, at Georgia Tech then what we're
trying to do is to form an integrated center or integrated approach to cancer by leveraging
the strengths that exist at Georgia Tech in these disciplines out here, computer science,
mathematics, material science, and so forth, and try to integrate that in with cancer molecular
biology.
That's just the idea of opening these communications between these different areas.
We just are starting this integrated cancer research center.
These are just some pictures of the PIs from the various departments to give the idea of
the breadth of this kind of endeavor.
And the other important component for us, since we're not a medical school, is where
are we going to get the samples and tissues to work on?
And so in our case, we established collaborative relationships with a number of medical centers
in Atlanta, and we're collecting tissues from these various centers being provided
to us.
I started off working on ovarian cancer because I was contacted by Dr. Benedict Benigno, who
was head of a group of surgeons in Atlanta who specialize in ovarian cancer, and they
did about one ovarian cancer surgery per day.
And so over a period of several years, we were able to collect a large number of samples.
Since then, we've expanded to other hospitals in the Atlanta area and others connected with
other surgeons who engaged in this endeavor in providing different kinds of samples.
We're also now getting some samples from TGen for pancreatic cancer serum studies, which
I'll talk a little bit about, although I'm going to focus primarily on ovarian cancer
in this talk.
So we've got the collaboration, a heterogeneous group of scientists and engineers here.
We've got the clinicians here and the source of samples.
So we think we've got enough to move forward with this.
This is just talking about the samples.
I told you the ovarian cancer, we've been fortunate, and we've been doing it for several
years.
So we now have over 1,500 ovarian cancer patient samples.
And so for ovarian cancer, for example, we get the tissue, we pay a surgical technician
who goes into the operating room who's charged with taking the tissue as soon as it's removed
from the patient, snap freezing it.
So we have high quality tissue that goes in.
We get serum, whole blood, urine, and ascites fluid from every patient if it's available.
This nurse is also charged with acquiring extensive patient history information on every
patient that's involved in our study, and that's all stored in a centralized database.
So those are the resources that we have.
And as I said, I'm going to talk specifically about ovarian cancer.
We already had an excellent introduction to that, so I'm not going to have to convince
you it's an important disease.
Probably the biggest problem with this disease is that it's asymptomatic, essentially asymptomatic
at early stages.
By the time a woman knows that she has this disease, it's almost always at stage three.
It's almost always metastasized, and therefore the prognosis is very poor.
It's very difficult to treat the disease at that stage.
So there's a dire need for better diagnostics for ovarian cancer, as for all cancers, and
better therapeutics as well, of course.
Now I'm going to talk today about three examples of this kind of collaborative research that
we do at Georgia Tech to give you just an idea of the kind of directions we're moving.
I want to put it within a biological context, and that context I'm going to use as metastasis
because that's been talked about earlier in this meeting.
I put Dr. Benigno's picture up here because he insists on it whenever I'm giving talks,
and there may be a friend of his here that will report back to him that I didn't put
it up.
So his quotation is that primary tumors usually don't kill people, metastases do.
I think that's a theme that has been coming up throughout this meeting.
So metastasis is really an important phenomenon.
I think we all know what we mean by that.
In the case of ovarian cancer, we're talking about primary tumors arising either in the
ovary itself, for example, as people are suggesting it might be low-grade tumors for
serous papillary arising there, and perhaps in the fallopian tube as well.
But that would be the primary site.
The metastasis we're talking about is the sloughing off of cells primarily from these
two primary sites and spreading through the omentum, through the abdominal cavity, and
oftentimes ending up at the omentum, which is here.
All right, so let's think about what are possible solutions to this problem of metastasis.
I think probably the best and most obvious solution would be if we could diagnose the
disease before it metastasizes, and we don't have to worry about treating it.
The treatment would be easy.
That is, if the tumor is confined to the ovary or even to the fallopian tube, it can be removed
surgically, and provided it has not metastasized, the disease is essentially 100% cured.
So if we can get early diagnosis at this stage, we could get rid of the problem of metastasis
right away.
The problem is if we move into stage three, and as I said, most women are diagnosed at
stage three and sometimes not until stage four, the tumor, the cancer cells that metastasize
all through the abdominal cavity is extremely difficult.
Basically it's impossible to get all of those foci out, and because of the heterogeneity
and so forth, chemotherapy is also problematic.
So I'm going to talk about our efforts to get a better diagnostic test, and this is
going to involve a collaborative project with an analytical chemist at Georgia Tech and
a computer scientist, and then I sort of am playing the role in the middle, providing
the tissues more than anything else, and show you how we've gotten on that.
So this is involving a rapid mass spectrophotometric metabolic profiling of ovarian cancer serum.
These are the PIs involved with it, Alex Gray is in computer science, he came to Georgia
Tech from the Jet Propulsion Laboratory, he specializes in analysis of high volumes of
data, he's into pattern recognition algorithms, which I'll get into in a few minutes.
Facundo Fernandez is the analytical chemist that does the mass spec, these are the students
and postdocs that were actually involved in the project.
So how I got involved in this, when I arrived at Georgia Tech about seven years ago, I was
reading one of the magazines where they describe the wonderful research that's being done at
that institution, I'm sure there's a magazine here that does the same thing.
They had a description of work done by Facundo, where he was using mass spec to identify fake
drugs, and you can see here he's using a technique called direct analysis in real time, he would
take the pill, put it within the beam of hot gas, it would pick up enough molecules off
the surface of the pill, he could run a spectrum, he could distinguish true, the actual drug
from counterfeit drugs based upon differences in the spectrum.
Not only that, if it was counterfeit, he could tell you if it was made in Shanghai or Indonesia
because it had unique profiles that were mapping to each of those places.
So because of that I thought, well, I wonder if we could do the same thing with serum from
patients, simply look at the profile and see if there's unique fingerprints of individuals
with cancer as opposed to normal, and whether we could use the same sort of approach.
So I just said a little bit about the direct analysis in real time.
Facundo was using that because it doesn't require any upfront preparation of your sample,
you can go directly from the sample, so it has some advantages, I guess, for diagnostics
because the ideal diagnostic assay should be non-invasive obviously, so we want either
blood, urine, or ascites fluid, or some sort of material that would not require surgery
to acquire.
It should be high throughput and low cost.
The cost of this assay is less than $0.50 per assay, of course it's expensive to buy
the instrument, but other than that it's fairly low cost, it's high throughput, and as you'll
see it appears to be highly accurate.
We decided that we were going to look in the spectrum range of metabolites, and so metabolites,
I like to define them simply as anything that's not protein or nucleic acid, throw everything
else into that bag called metabolites, and the reason we chose that really initially
was because there was less competition in that area, there was a lot of people doing
mass spec analysis of serum looking at peptide differences as possible diagnostic markers.
The other reason was the metabolome is much smaller than the proteome, so in theory we
thought that we would get better separation of the molecules as we did the analysis.
So it was really for those reasons that we chose it.
As a matter of fact, however, metabolites have been used as diagnostic markers going
back to the Middle Ages, and back in those days, here's a picture of what they used to
call the urine wheel, so they would get patients to urinate into a glass and then match the
color of it up to the color on this wheel, and that would tell them what the disease
was.
And those color differences in urine are really differences in metabolites.
So sometimes they would smell the urine, and you still do, you can tell if somebody
has disease, that's all metabolites.
And so the use of metabolites in diagnostics really has an extensive history.
Okay, in our case then, we would just simply collect the blood, let it clot, we took the
serum, we precipitated the protein out of it, we dried the sample down, derivatized
it so that it was more conducive to the mass spec, and then ran it into the mass spec and
got the spectrum coming out of it.
The data that came out of it then was shipped over to Alex Gray, where he used support vector
machines for his learning methods.
He was involved in pattern recognition, facial pattern recognition, where they would measure
the vectors from the distance of the tip of your nose to your eyebrow, and from your eyebrow
to your ear, and all that stuff.
Doing that, you'd come up with a pattern that's unique for each individual.
And then you go through an airport, and they take your picture, it goes into the database,
if it matches, then you suddenly have handcuffs on you.
The same kind of approach is used for retinal scans.
So we wanted him to just use that same approach in the analysis of the data that were coming
out.
When he gets the spectrum that comes out, where normally people used to look at just
the peak height or something, he looks at this as if it's a face.
And he's taking readings at various points along these curves.
So he's factoring into his analysis not just the peak height, but the shapes of the curves
and so forth get factored in as part of his database.
Using that approach, he comes up with greater than 20,000 features in a typical patient
sample that we run through.
So then the issue is that you go through it, and you run a learning set.
So you run maybe 50 cancer patient samples, 50 control, and you get the spectrum coming
out.
And he looks through it, looking for those features that are stable among the cancer
group, not found in the normal and vice versa.
And that becomes your model of the disease.
Now metabolites are highly variable.
Most of the features are highly variable because metabolic levels are affected by what you
have for breakfast and all kinds of things.
All of those variable features are thrown out.
We're only looking at those stabilized features.
So we go down from 20,000 features to something in the order of under 1,000, more likely in
the area of 100 features that are stable, and those become the model.
Then once you have the model, acquire your data set, you build your model using the support
vector machine to separate the values from the controls versus the cancer.
That model is stored in the computer, and then you run a test set.
And we give him 100 or so samples of serum from patients, not telling them is it cancer
or not, and asking them to tell us how do you classify these things.
And that's how we test the validity of our model.
So on the initial roughly 100, I think it was 94 or something, patient samples that
we ran through, he came back and he didn't know what they were.
We put the labels on them after.
You can see that with nearly 100% accuracy, he was able to separate the cancer samples
from the normal.
There was...
This is unpublished.
We just got this data last week.
This is prostate cancer.
You can see on 140 samples that we ran through, again, in this case, 100% accuracy.
Normal controls from the test.
And this is, if this pulls off, this is going to be very good as well, because the current
test of prostate cancer is not very good.
So everything looks good.
I don't know if I put up here yet.
We just started doing a little bit of testing.
So we're very excited about this approach being used to diagnose if it looks very promising
at this point.
So that's one example.
And I'm really an advocate of saying that the best, from a pragmatic point of view,
the best way to deal with cancer at this point is to go with early diagnostics, because therapeutics
are problematic for many reasons.
All right, the next possibility, though, would be, we thought, maybe removing metastasizing
cells from the abdominal cavity.
And so this was a collaboration that was started with a material scientist, John Zhang, who
worked with magnetic neoplasm.
And I had a postdoc in the lab a few years ago who was working on ETHA2, and she had
worked on ETHA2 receptor for her PhD work, and so she was interested in looking at it
in ovarian cancer.
ETHA2 is a receptor of heparin, and you can use heparin-magnetic peptides that are known
to bind to this receptor very well.
And so she was initially interested in it from a molecular point of view, but we thought,
let's just exploit this from an engineering point of view and see if we could hitch.
I should say that she found that this receptor is overexpressing ovarian cancer cells in
about 70% of our patients.
And so it was a fairly specific receptor.
We thought of the idea of hitching this magnetic peptide to the magnetic nanoparticle, and
the initial concept was, could we introduce these particles to the abdominal cavity?
They would clamp onto the cancer cells, and then could we sweep them up in a magnetic field?
That turned out to be naive, because the initial results looked good.
This is just showing some in vitro work that was done before, where we show the magnetic
nanoparticles are labeled in red here.
These are ovarian cancer cells.
So this is putting the magnetic particles in without any YSA peptide on them, so there's
no specific binding, but when the YSA peptide is on the magnetic particle, we get binding
to the cancer cells.
When we put human ovarian cancer cells in the abdomen of a mouse, and we put nanoparticles
in and let them incubate for a while, we just brought a high-powered magnet up to the abdomen
of that mouse, and you can see the ovarian cancer cells were tagged with a green fluorescent
tag, and you can see that the cancer cells came up to the magnet focus.
So we thought, well this is great, that might work, but the problem ended up that cancer
cells that had attached to many membranes and many tissues, the magnets went there as
well, and we couldn't get them out.
So the patient would start to be loaded up with magnets, and so that wasn't going to
work.
So then we had the idea, well maybe we can do something like a dialysis-type system,
where we could perfuse the patient, the fluid would come out, we'd have a chamber where
we added the magnetic nanoparticles, they'd go through a magnetic field, pull the cancer
cells out, and then recirculate them.
And we thought that also might be desirable because we would be not pulling out immune
cells, cells that might be in the patient actually being used to fight the cancer itself,
we'd be specifically pulling out the cancer cells.
And so here's a movie, this was, then we started doing this in a model system on the
mouse, this is a sketch of what was done.
You're going to look at a picture, cancer cells are going to be labeled again with this,
they're going to show up as yellow, they're tagged with that green fluorescent tag, you're
going to be looking at them flowing through the tube, when the magnetic field is placed,
when the magnet is placed in the area of the tube, you're going to see the cancer cells
come over to the magnet, which is sort of proof of concept that the system is working.
Okay, so there's the, you can't really see the individual cells, you apply the magnet,
so cancer cells are coming over to the magnet, so in theory then we should be able, this
proves that we could, in theory, pull them out in that trap.
Then we remove the magnet, the cells start flowing again.
So everything looked good at that point.
We then went on and took a CITES fluid from human patients and just simply showed that
using the magnetic particles we could enrich ovarian cancer cells after treatment with
these magnetic nanoparticles.
So we don't have a device large enough where we can do the dialysis type thing on humans
yet, the engineering firms want a million dollars to build a prototype.
So I'm going to have a payload pass around after the talk, we're looking for investors
that want to contribute to this.
We're stopping that, we've done a number of experiments showing that we can reduce the
frequency of metastasis, that's why I was asking if your model metastasized, because
we could use it in the system to see if we could reduce metastasis.
But if we artificially inject cancer cells into the mouse and then have controls where
we treat and don't treat, we can reduce metastasis significantly with this model.
Alright, the third possibility of dealing with this problem of metastasis is to customize
therapies against metastasizing cells.
So I'm not talking about attacking the metastasized cells, I'm talking about the ones that are
in the process of moving to the secondary site.
Because once they get to those other sites, it becomes very difficult.
So, like earlier talks in this meeting, I talked about epithelial and mesenchymal transitions,
and this is a model that is being favored by Robert Weinberg, which I also believe a
significant number of people do, that the primary tumor is largely made up of epithelial
cells, cancer epithelial cells.
At the periphery of the cancer, these cells go through epithelial and mesenchymal transition.
They generate these mesenchymal-like cells that are really the metastasizing cells.
And these cells, there's a number of things that happen to cells as they go through EMT,
they lose adhesiveness, which is conducive to them sloughing off the primary tumor.
They separate from the tumor, they're highly mobile, they're highly invasive.
And they're quiescent, that is, they're not replicating, which makes them problematic
because most cancer therapies are targeting rapidly dividing cells.
And since these metastasizing cells are essentially quiescent, they're very resistant to standard
chemotherapies.
And so, unfortunately, what we see is a patient will get treated with cisplatin or something
like that, you'll see the primary tumor shrink, and the patient will become very excited.
But the problem is that these mesenchymal-like cells have not been touched, and so you'll
get recurrence later.
These cells will spread to elsewhere in the body, in our case, to the momentum largely.
And then they convert, they go back through a process called mesenchymal to epithelial
transition.
They convert back from mesenchymal cells back to epithelial cells, and they're going through
rapid division again.
Okay, so we're interested in how to attack those mesenchymal-like cells to try to prevent
them from metastasizing.
And there's two basic ways you can do it.
One is through forced screening.
You can get from various sources.
Drug companies have thousands of these experimental drugs from NIH.
You can apply, and they'll send you down thousands of experimental drugs on these plates.
And you can take your mesenchymal-like cells and just run them through all of these drugs
and see what kills them.
And out of screening, around 1,000 compounds, we've come up with around 10 that appear to
be effective at killing these mesenchymal-like cells.
In many cases, how these drugs work is not known.
They don't even know what the targets are of them.
So we collaborate with a computer scientist who does structural predictions, and we give
him the drug and then ask him if he can give us an educated guess as to what the targets
might be.
And then we can go back in the lab and test his hypotheses, and then see if we can even
tweak the drug.
So I'm not going to talk any more about this.
It's kind of unimaginative, and it's kind of brute force approach to the thing.
Another strategy that we like better is a targeted therapy that would induce the mesenchymal-like
cells to convert back to epithelial cells, which render them sensitive to existing therapies
like cisplatin and so forth.
So we want to induce mesenchymal-to-epithelial transition as a precursor to applying chemotherapy
treatment.
So that brings me in the topic of regulatory RNAs, because the approach we're taking to
try to bring about these mesenchymal-to-epithelial transitions is to use these regulatory RNAs.
You saw a picture of a movie earlier of how siRNA works.
I'm going to show my movie, too.
So this is a risk complex.
Here's an mRNA.
So in the case of the siRNA, the homologous pairing occurs here.
There's enzymes in this risk complex that cleave the messenger RNA and then exonucleases
to the mRNA.
So this is how gene knockout is brought about through siRNA.
MicroRNAs are like siRNAs, except they're endogenous.
SiRNAs are usually something that we synthesize to have homology to our target mRNA.
But endogenously, as normal parts of our genome, we have microRNAs that are small-range reported
RNAs.
In plants, they work by almost nearly identical or an identical mechanism, because in plants,
the microRNAs have 100% homology to the mRNA targets.
So they behave just like siRNAs.
They bind to the risk complex, and there's cleavage.
That's the movie I showed you.
In the case of mammalian microRNAs, the homology is only 100% homology, 100% prostate-based
pairs of that.
And there's an area where the cleavage would normally occur is completely non-homologous
to the mRNA.
This prevents, then, cleavage by that same mechanism that's used to cleave by siRNAs
cannot be used for these microRNAs.
So you don't get cleavage of it.
You get binding to that mechanism, resulting in blockage of translation.
As we saw earlier, it could be explained why they're not getting translated.
In many cases, though, the mRNAs that are not translated that are blocked get transported
to P-bodies, where they get degraded by hexamutases.
And the P-body, that's why I was asking about that correlation.
So at any rate, there's a way of inhibiting the use of these microRNAs.
Now, I'm just putting it up here.
The human genome is believed to encode over 1,000 microRNAs.
They're predicted to target at least 60% of the human genes.
And so, you know, how do we know what genes are being targeted by what microRNAs?
Well, you can build algorithms.
You can go with the simple thing, sequence homology.
This is that defense pair region.
Where does it hit?
It shows homology.
But the algorithms are more sophisticated than that.
They take into account the shape of the RNAs and stuff.
And so there's a number of algorithms that predict the targets, the mRNA targets
and microRNAs.
And these are the three popular ones.
If you look at predictions from the three of them, there's a lot of overlap.
There's some differences between them.
But in general, they overlap fairly well.
You can also do in vitro things.
So we're planning to use these microRNAs in particular cell lines.
So we can go in to that cell line.
We can take the microRNA, transfect it into the cells, and then actually observe
what mRNAs are changing.
So we can compare what we observe to the predictions and see how well the predictions
are doing to predict what's being changed in the particular cell line we're studying.
In our particular case, we found that the Miranda predictive algorithm was the most
accurate in predicting.
It's still not 100%.
It's up around 70% overlap.
That is, we observed what was predicted about 70% of the time.
So we are interested in that because I have a bioinformaticist in the lab.
Chris Hill, in particular, is interested in this relationship between microRNAs and what
they're targeting.
So you can take microRNAs and you can group them into families based upon their sequence
homology, which you did.
Then you can take the predictions of what genes are being targeted by these microRNAs
and look at those from the perspective of networks, of functional networks.
So the interesting thing about microRNAs is that they target hundreds of genes.
Each microRNA targets hundreds of genes, but it's not haphazard.
It turns out that microRNAs target genes that are involved in generally similar functional
pathways.
And that makes sense from an evolutionary perspective because these things have been
evolving within cells for millions of years.
And they're there not as noise.
They're there to play some kind of functional role.
So it makes perfect sense that microRNAs would evolve to target genes that are functioning
related to one another.
And you can see that by overlaying these things with some of these functions.
In our particular case, so this is just saying different families of microRNAs, regular genes
involved in different forward-mated functions.
Some microRNAs, many of these microRNAs will target a similar function.
Some of them are very specialized, as you can see here.
Now, you've got to realize some of this is noise.
I mean, this all depends upon how we define functional categories.
We're doing that, nature's not telling us.
This is what we think the functional groupings are of these genes.
But anyway, we can get a pretty good idea.
And we're interested, remember, in epithelial and mesenchymal transitions.
So we're particularly interested in these microRNAs in here that are predicted to be
involved in that process.
So we picked one of those microRNAs, and we picked it because it's one that's changing
in expression in our patient samples when we compare microRNA profiles between normal
cells, precursor cells, and cancer cells.
And that particular microRNA we picked is miR-429, which is in a family called the
200 family microRNA.
And this is just showing you a close-up of the function.
So if we take miR-429 and say, what are the genes that are predicted to be regulated by
miR-429?
And once we have those genes, what is the goal enrichment categories?
And you can see that the most significantly enriched functions is epithelial and mesenchymal
transition for this microRNA.
So this is one that we wanted to work with because, remember, our goal is to try to convert
mesenchymal-like cells back to epithelial-like cells using some microRNA.
We're choosing 429.
I put 320 up here as a negative control because the top predictions for 320 has nothing to
do with EMT.
And so when we conduct these experiments, which I'm going to show you where we transfect
miR-429 into a mesenchymal-like cell, and S can convert it to epithelial, we want to
compare it with a negative control, which in this case will be 320.
All right.
So here's the experiment.
And this is a mesenchymal-like cell line, ovarian cancer cell line, called Hay 8.
And this is what it looks like.
Normally growing mesenchymal-like cells are spindle-like in appearance.
Epithelial cells are more combo-spindle-like in appearance.
And you can see when we transfected miR, these hay cells with miR-429, they converted
over.
This takes about, to get that kind of a shift, takes about 48 hours to see that complete
conversion.
If we transfect with 320 or make other negative controls like sprandle the base, we do not
get the change that these cells after transfection 320 still look like.
They're still mesenchymal-like cells.
Then we look at the changes on a molecular level, and we see changes in expression of
those genes that have been implicated previously in mesenchymal-epithelial transitions.
I don't know if I have a picture.
I don't have it up here, but we also check on these cells, their invasiveness, doing
assays to see how invasive they are, how migratory they are.
And everything fits.
That is, these cells are much more invasive and migratory than these cells.
Or vice versa, these cells are much more invasive, mesenchymal-like cells, because these are
the ones that are mimicking the metastasizing cells, and we're converting them back to
epithelial-like cells.
These epithelial-like cells, as we were going to predict, are more sensitive to chemotherapy.
This is cisplatin, and we're looking at these mesenchymal-like cells here.
This is compared to another cell line, this epithelial one, and there's a difference
in sensitivity.
If we take the hay cells, the mesenchymal-like cells, and transfect them with a negative
control of that 320, we get no change in sensitivity, still highly resistant to cisplatin, but the
ones that were transfected with miR-429 are now sensitized to cisplatin at about the same
level as other cells that look epithelial-like.
We can switch this cell from mesenchymal-like to epithelial-like with just your single microRNA,
and it acquires all of the properties of epithelial-like cells, including sensitivity to chemotherapy.
All right, so all of that looks good in vitro.
The problem is, to ever be able to move this to clinical practice, we have to have some
way to deliver these things, and that brings us back to knocking on the doors of the engineers.
So this is a series of studies that have been done with Andrew Wine, who's a material scientist
in the chemistry department, and he works with core nanogels, hydromanogels, as he calls
them.
And I know that there's been a number of talks here talking about different kinds of nanoparticles
to vector small RNAs, and I'm open to all of them.
It's just that Andrew is next door to me, that's why I'm using nanogels, but I have
no vested interest in nanogels, and we're looking for any kind of vectors that might
be good.
But we started off working with these, and he can adjust the pore size and so forth,
and we did all that optimization to try to adjust the pore size so that these microRNAs,
when they get to the cell, leak out at a slow rate, so that we have a window period where
it's converting its cell flow rate, which in theory sensitizes them to the chemotherapy.
So again, since the targeting that we used for the magnetic nanoparticles worked well,
we thought for proof of concept, we'd try pitching the same peptide onto these nanohydrogels,
load them up with, and I'm going to show experiments with sRNAs, because that was the first experiment
we did, but in theory you could use for microRNAs as well.
Target the cancer cells, and then see if the cells release the product.
Alright, so the gene that we're targeting in this particular experiment, I'm going to
talk about is EGFR, and the reason why is because there was previous work published
showing that inhibitors of EGFR increase the sensitivity of ovarian cancer cells to the
same therapy.
And so the idea is that EGFR, if it's overexpressed, ultimately, at the end of this pathway, it
suppresses apoptosis, and it makes the cells less sensitive to DNA damage, so they don't
go through apoptosis after the damage.
So since that was shown previously, we decided to develop an sRNA targeting EGFR to use these
nanoparticles to deliver to the cells.
So this is just showing, these would be the nanoparticles, and this is showing the cancer
cells, and this is showing the fact that we can get the cancer cells to attach into the
engulfed binding cells.
After this experiment was done, we looked at percent of EGFR expression, so this is
the YSA alone, YSA on the nanohydrogel, but no sRNA inside of it.
And it has an effect compared to untreated, so just the nanogel alone is doing something.
This is the nanogel without YSA on it, and here is the nanogel with the YSA peptide in
addition to having this sRNA against EGFR.
You can see we've got about, you know, over 80 or close to 90 percent knockdown with these
particles in detail.
And then, after we treated them that way, then we went and tested their sensitivity
to drugs.
This is showing doxetaxel.
We did the same thing with cisplatin.
And so you can see that the controls, the ones that were untreated, even at fairly high
doses of doxetaxel, their cells were laughing at us.
They're not affected by this drug at all.
But even cells that were treated with very low levels of these nanoparticles would show
significant increase in sensitivity to the drug.
So the concept seems to be working, at least in principle, the idea that if we can knock
this particular EGFR down, we can sensitize the cells to standard chemotherapies.
And we think we can take the same approach with microRNAs, that is that we can use these
particles to deliver RITR429, convert those cells from mesenchymal to epithelial, and
increase their sensitivity to cisplatin.
So we just received a grant from NIH to do this now in animal models.
So we'll see if it works there as well as it does in neutral.
Okay, so I went through and I gave you three examples of results that came about uniquely
from collaborations between people that normally would not in the past have collaborated.
And we wouldn't, as molecular biologists, we would never have been able to get any of
these things to work without these other guys that were working with us.
So I think that this is a good demonstration of the utility of this kind of integrated
approach that I think is going on in this campus as well.
So I'll just end here.
These are the people that I talked about.
So for the analytical chemists, Mr. Congo, and I talked about John Zhang.
These are both people in the biochemistry department, as well as Andrew is also in the
biochemistry.
Alex Gray is in the computer science department.
I'm in the biology department.
This is my lab.
Ken is the guy who did all the work on magnetic nanoparticles.
He's got a startup company now, and that's why I'm going to pass the hat around and give
him money for him to get money to create a device that's large enough to start phase
one project.
Roman is involved in the microRNA project.
Roman is in check.
Chris was involved in bioinformatics of the microRNA targets.
And then this is my wife, who works with me.
She's in the audience, and she holds everything together.
She insisted that I put a star on hers because she's much more important than me.
All right, thank you, and I'll take any questions.
Questions from the audience?
That was a great talk.
So is there something special about microRNAs versus any other kind of thing that would
make them a better target for delivery via this mechanism versus lots of other drugs
that you'd like to get?
The way I think about it is that microRNAs have been evolving for millions of years
figuring out what targets are important for our ability to transition.
We can go ahead from our own point of view and say, oh, I think snail or twist is important
in going in.
And snail and twist are genes that are regulated by these microRNAs, but there's a lot of other
Now, an interesting thing is that if we take some of these sRNAs against snail and twist
and transfect them independently, we do not get the mesenchymal epithelial transition.
So we don't know what the magic bullet is to get that switch, but microRNA does.
So I think we can actually use microRNAs to teach us what might be important targets,
because they've been figuring it out over millions of years.
So that's why I like them, because I think that they've already figured out how to control
processes, but we don't know how to guess it.
And along the same lines, presumably you show there are families that seem to cooperate
to similar things.
Have you combined any combinations, especially in cases where they have different sequences?
And some of them actually have, sometimes it's in different families with quite different
seed sequence, but we get the same genotypic change.
We haven't combined them, but we've tried, we've started using some of these other ones,
seeing which ones will give us the conversion.
They all tend to give the conversion, maybe not as intense.
429 so far is the best one for us.
That is at 48 hours, we get very sharp demarcation.
The other ones show a switch, but it's a little bit murky, which makes me think in
the real world that actually might be important.
We just don't realize the significance.
For us, the most important thing is the ticket over.
But maybe in reality, we don't want to do that.
I don't know.
Dr. McDonald, I have a quick question about the mass spec metabolome project.
I'm just wondering, I know it's high throughput, but have you ever gone and actually assigned
identity to each peak?
That's what we're trying to do now.
There are computational ways that we can narrow down to those features that are most important.
In the case of prostate cancer, we know now there's about 20 metabolites or 20 features
that give us essentially 100% of our accuracy.
Now the big thing is to go in.
You can make guesses just on mass charge, but to prove it, you have to go in with the
compound.
The company Metabolome has invested millions of dollars to synthesize all these different
metabolites, so they can go in quickly and identify them.
We don't have those resources.
So far, we don't want to give our stuff to them yet.
But you're right.
That's what we want to do, is to identify what those things are.
Because then you could come up with some kind of ELISA-type test that could be used in a
clinic.
This would be a discovery-based kind of approach, and that would be actually a clinical application.
Great.
Is there a sense of what the chicken of the problem here is?
What really is maybe initiating the cancer in the first place?
Are these environmental factors?
Or is it the estrogen and the risk-ability that the patients may be undergoing?
What is really the problem?
Well, I mean, if you ask 100 people, you're going to get 100 different answers.
My background is in evolution.
I was a molecular evolutionist until 10 years ago when I switched into cancer.
But my feeling from an evolutionary perspective is that these changes are occurring all the
time, and that what's holding us in check is that we have surveillance systems in our
body that will kill any of the cells that go outside the realm of what we call normal
functioning.
So I think it's important for us to look at both.
As we get older, those surveillance systems deteriorate.
From an evolutionary point of view, you don't exist after you reproduce.
And after you reproduce, nobody cares.
Evolution doesn't care what's going on with you.
So the problem is we're all living too long.
That's why cancer's a problem.
So I don't think it's unnatural.
I think the idea of cancer being an unnatural thing is probably the wrong way to think about
it.
I think it's natural that the system is breaking down.
What's kept it in check before reproductive age, from the evolutionary point of view,
was to evolve these surveillance systems to get rid of them.
So maybe one approach to cancer—well, the biggest approach is to get early diagnosis,
I think, from a practical point of view.
But maybe we should be looking heavily at these surveillance systems, and if they are
breaking down, try to reestablish their functionality.
And that might be the best way to go.
The other way to go, I think we're going to find all these genes involved, and I think
that's interesting, but I think it's going to be very difficult to use that information
therapeutically, because I think the cell will always find a way around it.
As an evolutionary biologist, there's many ways for us to drive to New York City, and
I can blow up the main highway, but if somebody wants to get there, they're going to find
another route to get there.
And I think the cell is so integrated, highly integrated, it's going to be very difficult
to block all the routes to New York, and to solve cancer that way.
The best way to do it is to kill the guy who wants to go to New York.
I didn't know, now that you want all of the patients, all of the people who need to undergo
estrogen-based therapy, to just say, hey, it would be great if you could get some of
these animals to get stage 1, targeted as possible stage 1 ovarian cancer patients.
Yeah.
I think that's what will happen, right?
We're happy to collaborate, by the way, with anyone that wants to collaborate, because
we've got samples, but we can't do everything with them.
But I think in the long run, yeah, that's where we're moving, a patient will come in,
the samples will be taken, we've got a patient, and then that will be available for analysis,
and from that information, we'll optimize the therapy.
Any ideas on what switch, or what's unleashing these microRNAs at the point of the anti-transmission?
Well, no, I mean, it's probably, not really, but I mean, they're under epigenetic control
in all kinds, just like all other genes.
These things, many of them map within introns of other genes, so those microRNAs are under
control of the genes that they're mapping.
And interestingly, in many of those cases, the microRNA will regulate the gene which
is embedded within them, so there's this feedback kind of thing going on.
In other cases, the microRNAs have their own promoters, and they're completely mapping
away from the other gene.
So it's going to be very complicated, and it may lie, it may be epigenetic, who knows.
And they also affect epigenetic changes, as well, so there's all this feedback integration,
but that's why, when I start thinking about this, I think, hey, because how are we ever
going to be able to deal with this in any rational way?
You're targeting mesenchymal cells as epigenetics.
Are those the only, is that the only situation when a cell becomes mesenchymal, when it's
trying to metastasize?
Or are there any other processes in the body?
Metastasizing cells are, in general, they're mesenchymal-like cells.
So-called cancer stem cells are also mesenchymal-like cells.
Are there any benign processes in the human body that are also involved in mesenchymal
cells?
Well, they're very important in early development, of course.
I mean, that process, mesenchymal cells are what we all start with, and then differentiation
comes off of those cells.
So EMT is going on in early embryonic development.
That's what it evolved for.
And so what we're seeing now, again, is a breakdown of the system in people that shouldn't,
from evolutionists' point of view, shouldn't, from evolutionary point of view, shouldn't
be alive.
And we're seeing a breakdown of developmental systems that evolved initially to work in
early embryonic development.
Fantastic.
John, since all the science is exhausted, I'll ask you a question that's pertinent
to our campus.
You did all of this so quickly.
What were the sort of driver mutations for this to come about and put all this together
in the space of a few years?
That was tremendous.
I mean, my feeling, the older I get, I realize what makes things happen are people, right?
I mean, you can't have a proclamation from the president that says, we want an integrated
biology.
There has to be people like you that there's interest in.
And I think if you go over and just knock on doors, it's like seeing Facundo's work
on mass spec, and thinking, well, go over and say, you know, you're doing an account
of the drugs.
How would you like to do it with cancer samples?
So if you have the samples, then they're happy to do it.
And these guys are all very smart.
It only takes them about 15 minutes to get way beyond what you would think about doing.
And that's what we ended up doing.
And I think the key thing is having the samples there so that it isn't just talk.
I mean, I can go up to them, and if they want to do something, here's the stuff to go and
do it.
And I think their interest is there.
Just take some catalysts to want to go and make it happen.
All right.
Well, if there aren't any more questions, let's thank the speaker again.