Dan Gallahan, PhD

kettlecatelbowcornerAI and Robotics

Nov 7, 2013 (3 years and 5 months ago)


NBTS Symposium


of 12

Michael Yaffe

Dan Gallahan, PhD

I did want to say a few words about the symposium. I think when Dave asked me to help
participate in this, shortly after a large meeting we had organized, I said yes and sort of as I
reflected on it said oh, why did I agree to do anothe
r meeting again shortly after having
prepared for one. And actually the answer came to me last night at the reception, the
dinner last night where we heard some very powerful and inspirational stories from
survivors and family of brain tumor patients. So

that really brough
t it home for me and
again, it
to reflect on why we’re doing this
. A lot of times, especially

in the
research field don’t really have that sort of connection. So that was an important aspect of
it. But actually the job is n
ot that difficult. You can see this is really a stellar panel that
we have here today, that you’ll be hearing from. And as we sort of approach these
speakers who, believe me,
are very busy and
, they all v
ery willingly agreed to

this panel

And I think it speaks to the fact of not only their commitment to the field but also the
importance of this emerging field of systems biology and the hope that they have
Now defining systems biology is always a difficult job, I think be
cause it is an emerging
field an
d you will hear a number

definitions of that.

I think if you look across the titles
of the talks that you saw, what we have in the program, I think that sort of defines what
systems biology encompasses, from basic under
standing of the molecules involved in it,
on through the system, processes of the cell on through, looking at a higher level in terms
of the response to various therapies and drugs that can be involved in it.

So I think the spectrum that we have put toget
her for you today, I think will cover it in a
way that really sort of shows you the breadth of systems biology and how we can all bring
it together in hopes of a greater understanding of cancer

and ultimately of treatment of

I think there’s no b
etter way to start this than with our chair, Professor Mike Yaffe. Mike
is a professor of biology at MIT after receiving a Ph.D. and MD from Case Western. And
he is really a pioneer

understanding some of the signaling processes and the cellular
ses that go together and looking at the systems level.

So again, I think we’ve got a great panel here, a number of speakers that are going to be
speaking today. I look forward to always learning more. I’ve heard a number of these
speakers but every ti
me I hear them, I glean more from the field. And I get more excited
about the opportunities that it presents.

So without further ado, Mike.

NBTS Symposium


of 12

Michael Yaffe

Dr. Michael Yaffe

I’m going to start by
thanking Dan and David for the invitation to speak. As Dan

when I
got this e
mail from Dan and David saying you’re invited to speak, it’s not so much
an invitation as a command and so I simply saluted and said absolutely.

Dan, as I guess you gathered, Dan
as been a real pioneer at the NCI in pushing the
systems biolo
gy approach and we’re all very grateful to all the effort that he’s done. And
what I want to spend the first ten minutes or so of my talk telling you about is sort of a one
view of systems biology and trying to emphasize to you why I think it is that we n
systems biology types of cancer research if we’re going to understand and make advances
in the treatment of brain tumors. And it’s obviously because we want to find some way to
go from the diagnosis to the proper treatment. And I would argue that the

right way to do
this is by looking at signaling networks, for example, in genes and gene expression
patterns that we see in tumors, together with processes that we know are going to be
involved in the therapy, things like DNA damage and cytokine and immun
e responses,
cancer cell metabolism and particularly, signals that we think the tumor cell sees t
hat are
important for its maintenance

By way of full disclosure, I’m an SAB member

and stockholder in a company called
Merrimack Pharmaceuticals which is com
mitted to br
inging systems biology to bear

cancer therapy.

Okay, what is systems biology? It’s like the elephant to the blind men. If you ask ten
people what systems biology is, you’ll get ten different definitions but the definition that I
like t
he best comes from a quote from Craig Venter and he said, if we hope to understand
biology, instead of looking at one little protein at a time, which is historically how we’ve
done this

which isn’t how biology works, w
e need to understand the integration o
thousands of proteins in a dynamically changing environment.

And therefore, what we
hope to do is to take this sort of systems approach and apply it in a way that we can use to
identify better therapeutic targets or combinations of targets for the treat
ment of brain

Now the challenge really is how do we go from genetic mutations to therapeutic targets
and I think you’
ll hear after my talk, from Lynda Chin about the C
Genome Atlas
which has really done a fantastic job of identifying a wide v
ariety of mutations, genetic
mutations in cancer.

But the challenge remains how do we go from identifying these
mutations to figuring out what drug or what treatment we should give the patient when
they walk in the door. I think this is best emphasized i
n a quote from Mike Stratton who
commented that, really elucidating the fact that cancer is a mutator phenotype and so
somewhere between 1,000 and 10,000 somatic mutations in the genomes of most
cancers, including glioma.

And therefore distingui
shing which mutations are responsible
for driving the tumor versus which mutations are simply pasture

mutations is a great
NBTS Symposium


of 12

Michael Yaffe

challenge in some tumors, particularly

for example, medulloblastoma, this may be easier
where the number of mutations is smaller.

But in other types of cancer, particularly melanoma and lung cancer, the number of
mutations is astronomical, more than 100,000. And this, I think, is going to be a great
challenge to the field.

Cancer genomics as I think you’ll hear has been very succe
ssful in compiling a parts list
that is telling us what parts there are and what parts are broken in different tumor cells.
And we certainly need that research. But what I’m going to talk about is I think we also
have to have mechan
ism context and tempor

. When we look at this parts list, I
think it’s a great challenge

for anyone in the room to take a look at these bushings and
couplings and hoses and gauges and tell me whether or not this is a water purification
system or a dispenser for Coca
Cola or the best espresso machine you’ve ever seen. And
the real challenge is to take it from this parts list to understand how the cell works.

Now one approach that’s been done is to look at RNA expression profiling. That is to take
a tumor, look at wh
ich RNAs are expressed and try to make some claims about how to use
this RNA expression data in order to successfully predict therapy and I think we’ll hear
quite a bit about that, I anticipate, from Andrea Califano. And I think this has real
potential bu
t so far, it remains unclear whether this will or will not work. So I think this is
still an open question.

What I would argue in this picture that I’ve taken from my colleague Bob Weinberg is that
really what dictates the cellular response of tumors to
different types of therapy is
signaling. It’s the way in which the parts encoded by the genome and expressed at the
level of mRNA are actually wired at the level of biomolecular circuits which are built from
proteins. And so, in the example that I’d like

to show you, I’m going to consider for
example a cell shown here in which it sees various cue
s. For example, TNF and EGF or

insulin or in the example I’ll show you later, chemotherapy or radiation.

And somehow
these different cues

get interpreted by the

molecular machinery of the cell in order to
predict the response, will that cell survive or will that cell die. And what we’d like to
understand is how is it that we go from these cues into the responses and the connection I
will argue is through signals

Now you can make a nice analogy between molecular circuits and electrical circuits and in
electrical circuits, it’s a straight
forward question to ask given a variety of inputs, what
will the output be. And the way that we, I think, would all agree to
do that is to measure
what currents and voltages are present throughout the electrical circuit. And we can do the
same thing with biological circuits in which case the currents of signaling for example are
protein modification, things like phosphorylation
. And we can think of enzymes that are
NBTS Symposium


of 12

Michael Yaffe

responsible for this, like protein
kinases, as wires that essentially carry the current of

Now in order to understand what signals are present in the cell and how they get integrated
to control the respon
se, what we really need to do is to have some kind of metaphorical
meters that we can use to measure in a time
dependent manner what’s happening with the
signals within this molecular circuit. We need some kind of a volt meter that we can use
to measure e
xactly what’s happening at the systems level in the cell. Now for electrical
circuits, there’s a good analogy. It’s called a probe card or a bed of nails tester in which
you place the circuit in the center of this bed of nails tester and it tells you at
hundreds or
thousands of nodes what the voltages and currents are. And we need to be able to make
those same kind of measurements, I would argue, in tumor cells if we’re going to predict
what tumor cells are going to do in response to treatment. The ulti
mate goal is based on a
fundamental premise which is that the cancer response to treatment is governed
by a
variate state of the signaling network.

So the different cues, for example radiation or
chemotherapy, become translated into different respons
es. We kill the tumor cell. The
tumor cell proliferates despite our therapy. We see invasion or the tumor cells differentiate
into a less aggressive form because the response that we see is some function of the signals
that we measure. So our goal, I w
ould argue, should be at least in part, to measure the
appropriate signals, measure the responses and try to decode the information processing
algorithm that converts those signals into the responses.

Now one thing that we’ve learned from engineering sys
tems is that the way to do this is
through complex mathematics, using formal mathematical models in which the parts are
represented in a universal framework and the dynamics of the signaling can be interpreted
in terms of mathematical representations. Now

even though this looks, I think, to most
people, certainly to many people who’ve trained in classical biology as somewhat scary,
I’m going to try to convince you that it’s really not and there’s a wide variety of
approaches that you can use.

Now the typ
es of modeling that we’re going to do depend on the types of data that we have
and the kind of questions that we want to answer because the types of models you can
build vary from very specific models based on, for example

differential equations in
we think we understand the mechanisms, to very abstract models, the type that I’ll
talk about in the second part
of the

talk which are really based on simply trying to
understand the relationships between signals and responses. And we can go all the way
rom very detailed mo
dels through logic models like B
oolean models and inference
models like Bayesian networks and mutual information content to these types of
relationship models.

Now one of the types of models that I’ll talk a fair amount about in the se
cond half of the
talk is an abstract empirical model in which we simply

we don’t make any assumptions
NBTS Symposium


of 12

Michael Yaffe

about what is happening in the network and we simply measure as much data as we can.
And then we use, for example, statistical approaches, in order to
try to relate the data that
we’ve got to the responses that we see. This is the most abstract type of model. And as I
said, we can build models if we think we truly have prior knowledge, that can capture
what’s happening to the cells, we can use much mor
e specific mechanistic models.

But I’m going to show you something that’s emerged for example, from using these
abstract types of models. As I said, it depends on the philosophy that you want to use
which is a function of the data that you have, as to wh
ether we can use a theory
model in which you have a significant amount of prior knowledge or a data
driven model
in which you simply take the experimental data and use a variety of different approaches
in order to let the data tell you what’s going

Now it’s a toss
up, of course, between complexity and detail. When we study single
proteins in great detail, something that
certainly did when I was a graduate student, we
get a great
deal of information at the mechanistic level about what that pro
tein does, shown
here in green, but we have very little predictive power to predict what’s going to happen in
the cell, as shown here in red. And of course, the more things we study, the less we
understand in mechanistic detail but the more our predictive

power gets up. And at the
moment, I would say as far as dynamic signaling data goes, we’re somewhere over here.
We’re certainly nowhere near the types of analysis that we can get from things like
sequencing data. But we’re a little better than
at signal proteins alone

Now part of the problem is that even though we do systems biology, I would argue that
systems biology is still quite data poor. That is things like DNA sequence analysis or gene
expression profiling, we can get a very large amou
nt of coverage. However, when it
comes to really understanding what’s happening within a single cell, I think that we’re
limited in this approach. When we start to look at signaling networks, for example, or
pathways, what we have then is data which is i
solated in little pools. We know a lot, for
example, about the
MAP kinase pathway
and a lot about the Delta/N
, but we don’t really understand how those pathways communicate with each
other. And so we have pools in which we have a lot
of local information but the local
information is not communicated effectively in our understanding between the different
pathways that we study. And this, I think, is really the challenge.

Where I hope we will go in the future as shown in these dotted l
ines, that is at the moment,
for example, static data. By static data, I mean we pick one protein. We measure it at one
point in time. That’s the kind of thing you can do in a brain tumor biopsy. You can
biopsy the tumor. You can measure one protein o
r a set of proteins and that static data, I
think we can certainly do more of than dynamic data at the moment. But

power is less. Dynamic data, by which I mean we’re going to do something to the cells or
the tumor and measure what happens
as a function of time. That data, I think, is much
NBTS Symposium


of 12

Michael Yaffe

more effective at being able to predict what’s going to happen but

we’re limited in our
ability to do that. And I hope in the future both for static and dynamic
data that

we can
move the level of complex
ity from where it is to a higher level.

Now at MIT, we sort of take a four
part approach to systems bio
logy and it’s largely based

being able to make very accurate measurements of as many things as we can measure,
using things like microarrays and hi
throughput Western blotting and imaging. And
then our goal is to take the measurements we’ve got and mine this using
proteomics and
genomics approaches of the type that I think you’re going to hear about in some of the
talks that follow. And then from

this data construct models of the type that I know
Thomas Deisboeck will tell you a little bit about, mechanistic models or biochemical
models or network models, that in turn we can use to predict what we think is going to
happen to, for example, a tumor
that’s treated with a particular type of therapy. And you’ll
hear about some of the experiments and treatments that we can use based on these types of
models, I think, in Stuart
’s talk.

So the idea is to use mathematical analysis to elucidate h
ypotheses and facilitate
predictions and then perform system
wide experiments in which we not only gather data
but we manipulate the system and we see whether or not the model predicts what we
observe. I would argue that a model has to be both insightful
and predictive. It has to be
insightful in the sense that it has to tell you something new that you didn’t know before
you started. It has to be predictive in the sense that it tells you which patients are likely to
benefit from a particular treatment an
d predictive in the sense of identifying new targets
that perhaps we hadn’t thought about before we did the work. A model that simply
explains the data that we

already gathered is not, in my opinion, a very useful model.

Now I’m going to give you one
concrete example. Everything I’ve talked about so far has
been in response to Dave’s request that I sort of give you an overview of systems biology.
And now I’m going to give you

a very short synopsis of one example of how you can use
this type of system
s biology approach to improve, I would argue, the treatment of cancer.
And I’m going to focus not on brain tumors but on breast cancer because that’s what the
work was done on. And the question that we set off to ask was could we examine, for
example, co
mbination chemotherapy using DNA damaging agents and signaling pathway
inhibitors in order to improve the treatment of breast cancer. And we focused on one
particular type of breast cancer called triple negative breast cancer.

Now breast cancers are cate
gorized on the basis of gene expression profiling into three
subtypes. You can categorize it in as many as 18 but I’m going to take the simple view
that there’s three types, and the three types include luminal breast cancers, HER2
overexpressing breast ca
ncers and a type called triple negative breast cancers. So most
breast cancers express either estrogen receptors or progesterone receptors or they
overexpress the HER2 oncogene. And a group of tumors, about 15 to 20 percent of breast
NBTS Symposium


of 12

Michael Yaffe

cancers don’t expres
s either the estrogen receptor or the progesterone receptor or
overexpress HER2, three negative things and they’re called triple negative breast cancers.
They are not homogeneous. It’s a heterogeneous population. But about 45 to 75 percent
of these over
express the

EGF receptor.

And so we set out to ask could we use kinase signaling pathway targeting, together with
combination chemotherapy in order to enhance the response of those triple negative breast
cancer cells which have the worst prognosis to trea
tment. And the way we set out to do
this was to combine various types of DNA damage, ionizing radiation or

etoposide or doxorubicin
, with a variety of

specific pathway inhibitors.
And what made this different from a systems

point of view was instead of simply adding
the two drugs at the same time and asking what happened, we asked what would happen if
we varied when we gave one drug compared to the other or we gave different doses of the
drugs at various points in time.

nd one story that I’ll tell you about has to do with doxorubicin and an EGF receptor
inhibitor called erlotinib. Now when we started the work, everyone who was familiar with
triple negative breast cancer cells told me, Mike, this is a complete waste of ti
me because
not only has this been tried in cancer cells in culture, it’s been tried in patients and it does
not work. If you measure cell death in response to treating cells with the EGF receptor,
you get a little bit of death. Doxorubicin which is a dou
ble strand, break inducer gives you
a little bit of death and you combine the two and you get just a tiny little bit of increased
death. Nothing to write home about. It’s better than either alone but not much.

But what we discovered was in fact we could

increase the amount of cell death by 500
percent. We could dramatically impact these tumors simply by altering the time between
when we gave the EGF receptor inhibitor and when we gave the chemotherapy. If we
waited four, eight, or 24 hours, we got a pr
ofound increase in our ability to kill these cells
compared to treating them with
doxorubicin alone. If we
the order, if we
damaged their DNA and then we blocked the EGF receptor inhibitor, in fact we made the
cells resistant to chemotherapy. It
’s specific for the type of tumor.

So this shows you the type of response you get

now on all of these plots I’m going to
show you, we’re treating with the EGF receptor or erlotinib or the DNA double strand
break inducer, doxorubicin and if we co
you’ll see a d/e. That’s doxorubicin
together with the EGF receptor inhibitor. And if we pre
treat, the first drug is shown here
and the second drug is after the arrow. So this is erlotinib before doxorubicin. Those
triple negative breast cancer cells
show a synergistic increase in cell death using this time
delayed chemotherapy.

If we look at HER2
overexpressing breast cancer cells, they show
just the opposite response. This is an antagonistic response. They become less sensitive
than if we hadn’t p
treated them. If we do this in the luminal cells, we get a slight
NBTS Symposium


of 12

Michael Yaffe

increase but that turns out to not be synergistic compared to what we see in the triple
negative cells. And in normal breast cells, we see very light apoptosis to begin with.

What’s th
e difference? What’s happening here? And so we set out to measure, as I
mentioned, as many of the pathways as we could that connected the DNA damage
response to the responses that we were seeing, the block in proliferation and apoptosis and
so everything

you see here in white is something that we could measure. And one post
doctoral fellow in the laboratory over the course of about six months made all of these
measurements you see here for 35 signaling proteins in three different cell types under all
those different conditions, using for example, high
throughput Western blotting that’s
shown here. So every one of these panels, this shows you, for example, ERK acti
vity in
response to EBP20 cells, in response to doxorubicin or EGFR inhibitors or various

combinations. And so we were able to get this big complex signaling data set that told us
what was happening to the signals under all of the different conditions. We could do the
same thing with respect to the responses. Mike was able to measure, for e
xample, how
many of those cells was surviving or dying by measuring ATP content. How many of
those cells were arrested in G2 or were arrested in S
Phase or in G1 or undergoing
mitosis? How many of those cells were dying by apoptosis? And how many of tho
se cells
were using autophagy, a mechanism in which the cell eats parts of itself in order to try to
stay alive in response to chemotherapy? And then we could use mathematical modeling in
order to ask how does those signals that we’re seeing relate to the

responses that we see.

Now we have 35 signals, so we’re looking at 35
dimensional space, and we use this
principal components partial least squares analysis to ask could we find particular
directions, principal components in which we were walking in that

dimensional space
to best capture the responses.

When he did that and the goal of course

is to identify what that information transfer
function is that tells us about all the signals he measured and how they’re related to the cell
fates. What came
out of it? He was able to build a model for those triple negative breast
cancer cells and the model

out to predict very well with two principal components,
the amount of cell death as well as whether or not the cells were proliferating or non
ferating. And it worked, as I said, quite well at being able to predict the amount of
cell death. So we could put in the values that we measured for the signals under one
condition and we could predict the amount of apoptosis and we could measure. I hop
e you
see here that it’s actually a very close correlation.

But then we could interrogate the model. We could say what’s responsible for the

cell death in a subset of breast cancer cells that respond to pre
treatment with an
EGF receptor inhibi
tor. The surprise was the single most strongest predictor turned out to
be caspase
8, a molecule involved in cell death. And you’ll notice, I hope, this blue bar
which represents the triple negative breast cancer cells is the highest for caspase

NBTS Symposium


of 12

Michael Yaffe

for the other cell types, the cells that were antagonistic or did not show synergistic
responses, those are shown in green and red. And I hope you see that caspase
8 was the
least predictive thing that came out of the model for those.

Well, this was quit
e a surprise because the traditional dogma is that DNA damaging
chemotherapy activates a death pathway that goes through caspase
9 and caspase

8 isn’t supposed to be involved at all in this death pathway. Caspase
8 is
involved in

death pathway
. And so it suggested that maybe we’d uncovered
something unique that you could do to rewire signaling in tumor cells using conventional
chemotherapy that’s already in the clinic by simply ordering the way in which we deliver
these drugs.

n fact, what the model predicted was caspase
8 was critical for the amount of
cell death that you would see in the pre
treatment condition only and if you co
treated the
cells, it wouldn’t make much difference. It was only in this pre
treatment condition
we would see this big difference in cell death if we had or didn’t have caspase

Mike tested that by using RNAi

to knock down caspase
8 and exactly as predicted, in the
triple negative breast cancer cells, if you eliminated that caspase
8 signal, y
ou lost the
synergistic effect of pre
treating the cells with the EGF receptor inhibitor. You had
essentially no effect on the cells that showed it an antagonistic response, suggesting that in
fact it was a new pathway that you had been able to unveil by
suppressing the EGF
receptor pathway in a chronic fashion but not an a
cute fashion. Does it work in v
This is a mouse model of breast cancer. What I’m showing you is the growth of tumor

it’s a xenograph model in which we treat the breast canc
er cells with a single dose of
doxorubicin. We get some decrease in the tumors but the tumors regrow over time.

If we co
treat with doxorubicin and the EGF receptor inhibitor, we get a little bit more of a
reduction in tumor size but the tumors pick up

and grow again. And in this experiment at
least, if we pre
treat with the EGF receptor inhibitor and 12 hours later, we treat the mice
with doxorubicin, the tumors not only show the initial response but they do not regrow
over the time course of the expe
riment before the vet said we had to stop the experiment
because the control tumors were getting too big.

Is this a cure for triple negative breast cancer? I would say not for all but for a subset of
them. If we look at
triple negative breast cancer
cell lines, what emerges out of this is
that there are four, only four of these triple negative breast cancer cell lines show this
synergistic killing if we combine EGF receptor inhibition with doxorubicin. It does not
correlate with p53 status and it doe
sn’t correlate

I’ll show you in a moment

with EGF
receptor levels. And so if we sequenc
e these tumors or we measure

EGF receptor
expression level, we would never have been able to predict which tumors would or
wouldn’t respond and that’s what I’m show
ing you here.

NBTS Symposium


of 12

Michael Yaffe

What I’m showing you here are those 10 triple negative breast cancer cell types and I’m
ranking them in the order of synergy to this combination therapy in which we pre
with the EGF receptor inhibitor. We allow the cells to rewire t
heir signaling network and
then we induce DNA damage by doxorubicin. So the dark bars are the most synergistic,
the white bars are the least synergistic. And you can see that the pattern of synergy for
most to least has absolutely no correlation whatsoev
er with EGF receptor expression level,
at either the RNA or protein level. No correlation. But if we measure the
of the EGF receptor in those cells as a measure of how dependent their signaling is, then
we find a very strong correlation.

Those cell types that are signaling strongly through the
EGF receptor are exactly those cell types that show this increase in apoptosis with this pre
treatment regimen and those are exactly the same cell types that show caspase
8 activation
as part of tha
t mechanism.

So to summarize what I’ve told you as an exampl
e of systems biology applied to

cancer is a way in which we can identify a subset of

I would argue of patients, certainly
of cell lines that would show enhanced response to chronic but n
ot acute suppression of
the E
receptor and the way it works is that under
l conditions

I haven’t had time
to show you this

the EGF receptor is driving an oncogenic signature through Ras that is
suppressing a caspase
8 pathway. So the conventiona
l chemotherapy can only work
through that intrinsic death pathway to cause cell death. We can rewire cells
therapeutically by using systems biology and pharmacology in order to suppress that
oncogene signature and unveil a second death pathway. So the co
chemotherapy can then be much more effective in killing these tumor cells because it can
use both the intrinsic and extrinsic death pathways.

So what I hope I’ve convinced you of is that we can use the conventional drugs and
treatments that are

already out there, if we just use them in an intelligent way, guided by
systems biology.

I’ll stop there. All the work that I mentioned to you was done by Mike Lee, a very
talented post
doctoral fellow in the lab with some help from the people you see

here. It
was a joint project together with our lab and Doug Lau
rger’s lab

and Peter Sorger’s
lab and

I really have to thank the ICBP Project at the NCI for funding this. Thank you
very much. I think there’s a minute or two left for a brief questi
on or two.

Michael Berens, PhD

Very exciting to take drugs as a way to provoke the system, Mike. Mike Berens from
. I was really impressed

that you had such a clear mirroring between the in
studies of the breast cancer cell lines and then t
hose were the same ones that went into the
in vivo xenograph model?

NBTS Symposium


of 12

Michael Yaffe

Dr. Michael Yaffe

That’s correct. Those were the BT
20 cells that we used in that xenograph model.

Michael Berens, PhD

That’s unique. I mean often we don’t find such good correlations
. Do you want to make a
comment about that?

Dr. Michael Yaffe

I think it’s really important to try to find the right way to do this with the right breast
cancer model. What the xenograph study shows, I think, is that there’s nothing special in
this part
icular setting about the stromal components and the other components of the
pathway. What really remains to be done, what we really need to do next is, of course, to
try various models of triple negative breast cancer cells in mice or in patients. At thi
point, I would say because we know

we were very selective in what we chose to really
devote our studies to but because erlotinib and doxorubicin are both approved for the
treatment of triple negative breast cancer, there’s no reason why we can’t go to
immediate clinical trial. We don’t have to file an IND. We don’t have to file anything
with the FDA. And I think we can do more mouse models which is worth it but certainly
our data suggests it’s worth a trial in humans.


Thanks, Mike, very
nice. I was also wondering

so you pre
treated with doxorubicin.

Other DNA damage chemotherapeutic agents or even radiation,

you see similar

Dr. Michael Yaffe

So we do see similar effects with camptothecin and we’re in the process of doing
experiments now with cisplatinum. We have not looked at radiation.


It may be a little different combination going through…

Dr. Michael Yaffe

Yeah. I think this idea that you can dynamically

rewire networks, that when we think
about combi
nation therapies for example, the idea that you can truly rewire
, how


in a way that you can then expound therapeutically. So you can sort of
state stacked cells. You can take the tumor cells and convert them all over time to a sta
that makes them uniquely sensitive to a particular treatment, I think has real allure and
we’re certainly pushing in that direction.

NBTS Symposium


of 12

Michael Yaffe


You’re doing it in sort of a general way with these agents but you could even see where
you could envision by

molecular engineering and really…

Dr. Michael Yaffe

So one of the things that we’ve talked about is doing a synthetic biology approach in
which we basically identify which signals are dominant and directly rewire those so that
only in those cells where t
hose pathways are dominant will that directly now activate a
death pathway. That’s exactly what we’re trying to do.


I have a quick question. Nice talk. I wish I had your slides actually.

So to be clear, you
get an increasing caspase
8 response to

erlotinib in some cell lines. So would you call
those cells on

Dr. Michael Yaffe

So it’s a complicated question, Vito. So what we found was that you could only induce
that caspase
8 pathway if you chronically suppress the EGF receptor.

If you acutely

it with erlotin
ib and then treat it with doxorubicin, you didn’t see that up regulation.
And we believe that these tumor cells, not necessarily just through this work but through
some other studies, we believe that the tumor cells
that we’re seeing, that are doing this are
the ones that are in fact on
addicted to the EGF receptor through the Ras pathway.
Oncogene addiction probably is a

it’s undoubtedly much more complicated than that
simple concept. But I think the concep
t is useful. So yes, I agree with you.