That everybody full disclosure we
actually went to grad school together and

so I go to flea most introductions.

In jail recounting of the theaters I play.

But if you are find yourself.

In a competition and

your detective is essentially the
equivalent of the catwalk this is the guy.

Who can do it.

But that being said I think the great
will give us an excellent presentation

he has a lot of awards and
honors today means that we're working for

the winner and I scaled one minute for
the final element ward and I and

I think and as you can see let's see if
we can get out there to as you can see

from his final there there is hopefully a
lot that is going to be appealing in terms

of chemical engineering concepts and what
I can say is that in terms of passing on

information to the seminar selection
committee pretty sure that what

won over the committee was this email that
I forwarded from the grave that was to me

that I sent off sounds good
yes the current mission

working on has done polling numbers
the own numbers partition coalitions

everything that chemical engineers
are totally enjoy reading to believe.

Miley days and that the real
smiley face is an invention so

with that all part of gray and
what we're going with your book.

All right thank you very much for that and
there a scene introduction and very happy.

Very happy to be here so thank you again
for the invitation I'm really excited to

share some of our more recent
work with you and just to get.

The out of the way and in the title
addressing decades old medical problems

with cutting edge chemical
engineering you may be saying and

I really just catering to the audience
this is a thinly veiled attempt given

all the chemical engineers and I want to
say that's absolutely not the case I'm not

bailing this is all clearly catering
to and pandering to the audience so

hopefully you'll enjoy the chemical
engineering approach we take to a lot of

these medical problems and so there's
actually a rich history of chemical

engineering in drug delivery not just with
what many of you may be familiar with that

with the polymer based drug delivery and
controlled release and nano particles and

materials but even going back into
the one nine hundred fifty S.

and sixty's some of the early
pharmacokinetic distribution

a lot of the seminal work in those fields
physiologically based pharmacokinetic

models looking at how structure
relates to distribution

was done from through a lot of chemical
engineers in that field and so

this is the vein that I'm going to come
to you and talk about some of our work in

quantitative systems pharmacology and
how we can use this to really drive

new medical imaging new molecular imaging
probes as well as drug therapies and so

our lab is both experimental
as well as theoretical so

simulations but
primarily we do a lot of experiments or

it's really the simulations that
are driving the experiments and

we kind of break down our research
group into two main areas one is

quantitative pharmacology and the other
one is molecular imaging agent design but

you can think of this is really
a multi-skilled problem and so

trying to understand how drug
properties whether it be

a small molecule whether it be a protein
or nanoparticle how those interactions

interactions way down here at the
molecular level impact the distribution

across all these different
lengths scales and time scales so

understanding the specific by
need interactions understanding.

Cotto interactions
nonspecific interactions with

proteins how that impacts with the biology
at the cellular and subsequent

level partitioning into different
compartments but then how it interacts at

the physiological level the tissue
the heterogeneity within the tissue and

how that translates into your organ level
distribution and finally at the system

level how this impacts plasma clearance
because every drug or every molecular

imaging agent needs to go from the site of
administration to the site of action and

really a molecule has no idea whether it's
a therapy or molecular imaging agent when

you inject it or ingest it it goes
based on its structure and so

the underlying theory of does that
developing better imaging agents and

better therapies is
really very similar and

it involves this scale as
well as both in length but

also time scale so the millisecond
interactions nonspecifically with proteins

all the way up to at the to see
over days weeks and months.

So I just want to start off with
thanking the people that do the work

every time I share this light I
realize they need to update it you

can tell it's extremely dated I look
very young and happy there and so

I need to put a new picture up there but
some of these people have moved on and

some of them are even
in this department but

I wanted to include them because all of
them were involved in some way shape or

form with some of the research can be
presenting today and of course the funding

source is several pharmaceutical companies
a lot of them don't want their name on

there for M.D.A. purposes but
Lilly is one of them a few others and so

there's been a lot of support from
the pharmaceutical industry so

I'd like to thank them and of course
government sponsors as well all right so.

We'll take a short break here and
reset power point.

But I'm going to move on wanted to
introduce the outline here that's true.

So in this one.

And we've already moved through
the introductions were making good time

I'm going to kind of lump
these next two together

in the talk here we go over biological
transport and simulations and

I'm going to lump that into one of the
first areas that I want to talk about so

I'm going to spend too much detail going
over the couple of P.D.E. models and

understanding all the transport parameters
I'll provide a brief introduction but

really what we're interested
in is applying those two

very difficult challenging but
solvable medical problems so

these are not esoteric simulations that
are just important to modelers but

these have real world implications and
impact on efficacy and so

I'll start with antibody drug conjugates
primarily with cancer treatment but

in the second half I want to talk about
two imaging applications one is in

developing imaging agents for diabetes and
the second one is in imaging for

screening purposes but throughout this
I'll be tying in the fundamental chemical

engineering principles that really
guide the research itself so

just to start off with the example
with antibody drug conjugates so to

remind everyone these are large proteins
antibodies one hundred fifty kilograms and

size the idea behind this is they
bind to cancer cells you can put

these stars represent small molecule
potent drugs that then bind the cancer

cells you can concentrate the drug on
these cells and selectively kill them so

you don't have side effects so it's really
linking these small molecule payloads to

the antibody for us it's academically
an interesting problem because we have

to understand and stimulate both macro
molecular distribution as well as

small molecules but from a medical
perspective for pharmaceuticals this is

a rapidly growing area there are over
thirty drugs in clinical trials

lots of next generation linkers and
payloads that are going in there

actually I didn't update this
slide since a day or two ago but

this is just changes been to more
approvals that just came out by the F.D.A.

a day or two ago however I want
to point out an important fact.

So those of you who may be familiar
with the drug delivery world so

one of the first tire seen
kinases hitters this is a big

breakthrough in targeted therapy was
Gleevec that came out around two thousand

since then this is a separate class of
drugs that I'm not going to talk about but

for comparison purposes it came out since
that drug came out there have been twenty

eight and I probably need to update
that number two new tires and

hitters that have been approved over
the past fifteen to twenty years

the first antibody drug conjugate
came out around the same time.

And it's been an active area of research
since that time there are have now been

only four new eighty Sezer or

C four Total eighty season even more
importantly only one of these actually

targets solid tumors that's the cat
Sila Targets Breast Cancer and so

despite all this development and this
investment it hasn't progressed as quickly

as some other categories so why is
that the case well this is something

that I've been continuing on my current
lab I started way back during my Ph D.

and a simply the majority of these drugs
that you inject never reach the tumor and

what does get there is very poorly
distributed so this is just an image

of a tumor where we forcibly tagged one of
the antibodies we've done a lot of work on

making sure we can do that without
disrupting the distribution begin C.

It's clearly not uniformly distributed
it's clustered around vessels and

not a lot of the drug makes it there.

So there's many cells in these tumors that
aren't receiving any drug Now this is been

a largely ignored in the pharmaceutical
area because you can dose your way out

of that so if you increase the dose
tenfold now you reach all the cells and

it's not a problem however with antibody
drug conjugates that's not possible

because the small molecule drug you have
on there is toxic enough that you can't

give a very large dose and in fact you
have to wait a long period of time

between doses as well so you end up
with a situation where you have.

Very heterogeneous distribution and
this is really under appreciated so

we want to explore this from the
simulation standpoint again I'm not going

to show any of the P.D.S. but you can
really break down any drug in any tissue

to several major steps and a classic way
of consolidating this is looking at which

are the rate limiting steps but every drug
I'm showing an antibody here Esteve flow

through the tissue extract as it across
the blood vessel wall it's a first

order mass transfer coefficient typically
diffuse Sometimes you have convection in

the interstitial M And then for eighty Cs
You have to be internalised integrated and

this is what releases the small molecule
drugs working against you are two types of

clearance one is from the blood itself so
you lose the concentration gradient within

the blood and in this case many drugs
will be metabolized within the tissue and

you actually have a local clearance so

this is kind of our overall framework
we've done this for both macromolecules

our idea is that will be able to take
any molecule whatever the structure and

predict exactly how it's going to
distribute we're very close to achieving

that goal I think it works a lot better
there's many in the medical community that

the don't believe that that's possible but

I think with a lot of our imaging we're
really approaching that goal and this of

course is it goes well beyond my work the
screws back decades what we're doing is

assembling a lot of the pieces together
in a comprehensive model and being able

to fill in the gaps in knowledge with some
of our imaging but I just want to bring up

we'll be talking a little bit about
small molecules that are released and

that behavior of small molecules is very
different you have membrane Permian billet

but again these are lipophilic molecules
in general so they do partition

the classic rock to know water partition
coefficient estimate how much gets into

the membrane they're very sticky they are
not very soluble so they stick to proteins

so we really have to understand how the
structure the lip of Felicity the number

of hydrogen bond donors except
the blocking the way impacts on.

All these rates of membrane permeability
partitioning as well as diffusion if it's

stuck to proteins or not and

really coming up with a comprehensive way
of making predictions counter to the macro

molecules which primarily diffuse in
the interstitial but we've done this and

we've measured a lot of these parameters
we do a lot of fluorescents microscopy in

a vital interval microscopy to measure
these rates in order to understand

how these molecules distribute you can
really break it down it is a complex

problem multiple linked P.D.'s but you can
break it down into about twenty parameters

that determine some of those are tissue
based parameters based on the physiology

are there are drug properties like
molecular weight lift of Felicity However

just to give you a sense of the difficulty
of the problem even if one of them is far

off that can completely change
your results and conclusions so

there's been a lot of data fitting
models I think what we really try and do

to differentiate our work is to make them
purely predictive because only then can

you have a falsifiable prediction that you
can go in and then measure experiment so

there's lots of models out there
where you can take data and

fit a line to it what we try and do
which I believe is much more challenging

is take a blank sheet of paper draw a line
on it give it to anyone who's going to

do experiments say if you take this
molecule with this structure inject it

your data points are going
to line up with this line so

it's a challenging problem but we've
been making a lot of progress in that so

we're starting to turn the corner
where we're then applying those models

because once you have something built like
that you can then design new molecules

based on your simulations so
this was a good way that we built

off of some of those P.D. models so we can
plug them in and we're really just filling

in some of the gaps where these parameters
this is too small to see but just

showing you there's multiple parameters
that go into these models that's great for

numerical simulations but what we like to
do is engineers we like to simplify and

come up with different regimes or
different space and so

we can take all these P.D.
ease and scale them and come up.

With some very simple concepts to at
least give us a qualitative flavor for

what's going on with these molecules and
again I don't want to belabor the point

I'm not going to go into too much detail
but some of the dimensions numbers that

can tell you a lot about how a drug or
distribute are shown here so

this is a dimensions number a vessel
depletion number if we think of a very

simple geometry think of a vessel like a
cylinder where you have a blood vessel and

then drugs are flowing in and extract as
a thing if you look at the blood flow rate

versus the extrapolation rate the rate
at which a permit ability to exit

if you have very high Permian billet E.

the concentration will drop along
the length of the vessel and

then blood flow in that tissue
becomes important but if the permeable

is very low relative the blood flow then
there is no concentration gradient and

you can get rid of that extra dimension
it just becomes a constant concentration

throughout the vessel the same thing goes
for abuse number this is a mass transfer

transport number we're looking at Permian
believe versus diffusion this tells you

the concentration drop going from
inside the blood to outside the vessel

if it defuses away very quickly relative
to Hermie ability you have a large drop in

concentration if not you have
equilibration with the plasma and

that's one misconception that often comes
up is it's easy to measure blood but

it is rarely the case where
the concentration where you want it at

the side of action is equal to the blood
and so we really have to understand all

these different not transport parameters
to know how much is there and finally

there's a couple different types of Dom
color numbers in this case we're talking

about reaction rates binding relative
to diffusion and I showed you that very

heterogeneous distribution well to jump
ahead if you look at antibodies there

binding rate there are a mobilisation
rate is much faster than diffusion and

so that's why you end up with this
very heterogeneous distribution.

But we do a lot of molecular Imogene or
multi-skilled imaging as well this

is a case we're looking at
histology slide we had that simul.

And we went ahead and said based on our
model let's take four different molecules

with different behavior that will each
have one of those rate limits steps blood

flow per me ability diffusion are binding
and would use four different floor force

and so we can track them all within the
same tumor and without going through it in

detail we can show that even within
the same tissue depending on the molecular

structure you can have all four types
of distribution within the tissue.

All right so I stepped over a lot of
the dimensionless analysis and P.D.

equations we can talk about that
more later if you're interested but

just to conceptually give you an idea of
what's going on for these eighty C.S.

there's a lot of interest in this market
it's a rapidly expanding market and

they typically don't do
this transport analysis for

their drug development show you
how important it becomes so

as I mentioned before there are multiple
ways of binding in the Fusion and.

The conclusion with antibodies is
that they buying the much faster than

they diffuse so as soon as it enters
it binds and the only way you

can get penetration deep into the tissue
is having another antibody come in and so

you actually have this what was
classically known as the bind site barrier

but really is a dynamic saturation
front slowly moves through the tissue.

So firstly all antibodies behave this way
but that's not a problem you just add

a higher dose again I mentioned
earlier the two limiting cases if you

have a dimension this is another dimension
this number if it clears much faster than

a penetrates as it leaves the blood you
have no more free anti-bodies to penetrate

the tissue suit will never reach all the
cells normally I say this isn't a problem

with antibodies there's some of the most
slowly clearing agents known they

often have half lives of weeks and so
you give them and many doses and so

eventually if you keep dosing you should
reach all the cells however with A.T.C.

is actually can be a problem because
they clear relatively quickly and

you have to wait a long.

Period of time between doses so

you actually can have limitations from
blood clearance but more importantly

you have to worry about local metabolism
or internalization This is of course.

Where if the rate at which these drugs are
entering the tissue is equal to the rate

at which there be metabolized even if you
have a constant concentration in the blood

you'll never reach all the cells and this
is very problematic for antibodies and

therefore eighty sees as well so this is
kind of a background let me jump right

into some of the more recent work we've
done with antibody drug conjugates So

we've taken this one drug that binds her
to which is overexpressed on breast cancer

it's the only drug that's F.D.A. approved
only antibody a drug conjugate that's

F.D.A. approved for solid tumors and we
use this as our model system we went ahead

we took a mouse tumors in a graph model
where we inject cancer cells underneath

the skin of an immune compromised
mouse and then we went ahead and

gave the clinical dose of that antibody
drug conjugate and we tagged it and

we could see that it only Reeses a few
cell Ayers outside of the blood vessel and

in fact is just change with
time this is maximum uptake but

after three and
five days it's frozen in time and

this is because of that
internalisation I talked about but

what's more important what we're looking
at is the number of cells so we're trying

to quantify the exact number of payloads
averred living person delivering per cell

So what's interesting is if you look at
the number of payloads delivered to these

few cells it is much larger than what
you need to actually kill those cells so

you're really in a situation where you
have over kill you have some cells a small

fraction that are receiving way more drugs
than what they need for cell death and

in the vast majority are around ninety
percent never see the drug at all and

so this is a major problem we know based
on how antibodies distributed this is

going to occur for all eighty C.E.O.'s.

And so we wanted to investigate this.

More deeply now I said I did
some of this work in my Ph D.

It kind of took a hiatus and did some
molecular imaging for my postdoc but

when I came back this again wasn't
really being analyzed in the field so

I thought maybe this really isn't that
important people have looked at it and

not causing any problem if you talk to a
lot of experts in the field they're saying

that it's not really an issue we've never
seen an evidence that causes problems so

we wanted to investigate that So
to come up with a model system to look at

we're going to spike in
some excess antibody So

to follow what we're doing here we have in
green we have this antibody drug conjugate

with the drug attached what we did
there's another clinically approved drug

drug Herceptin are trustees of is it's the
exact same antibody just no drug attached

so our thought process was if we can spike
in different amounts of trustees I'm abt

we can block some of those binding sites
and we can then titrate in the amounts and

control the heterogeneity
the spread of the A.D.C.

within the tumor and so to do this
we also want to make sure that we're

just looking ahead originating so we had
to find a tumor model that was completely

resistant to the antibody itself it
only response to the payload and

I'm going to go into the details but
from the immunology standpoint we had

to make sure that the I mean system
recruitment even in these new.

And K.
cells around we had to grow a larger tumor

A.T.C.C. is not an impact if that
interests you we could talk about it

more later but suffice it to say we used
a model system where the only effect

that we're seeing is from this payload and
so notice that we're when we spike in

the trustees and we're not increasing
the amount of payload the eighty C.

in the tumor all we're doing is spreading
it out and so first we want to show that

we could achieve this so this is in that
same model system where in red I forgot

to mention of the blood vessels
in green this is the tag drug at

levels that don't impact the distribution
and we can show that if we do a three to.

One ratio of the trustees are mad
we get better penetration and

then an eight to one ratio we get even
distribution throughout the tumour Now

these are actually fairly high doses of
trust to Zermatt but when this drug went

into the clinic there is no max or they
never found the maximum tolerated dose so

you can add very high doses of naked
antibody it's really the A.T.C.

that limits your toxicity so before
we did a lot of expensive experiments

with with our own mouse model system
we want to see is there any evidence in

the live sure that this matters and so
knowing how these drugs distribute We

wanted to look at some experiments that
have been run by a lot of companies is

changing the number of small molecule
payloads per antibody So if you follow my

logic here what we want to do is look at
situations where they change the dose and

the the DA or the number of small molecule
drugs for antibody in the same ratio

such that the dose of the small
molecule they're giving is the same.

But it's spread out as
conjugated to more antibodies so

if they do that we know the higher
antibody dose is going to penetrate more

but we're going to have the same
the same amount of delivery so for

instance instead of delivering
one meg per kid of a dar for

if you'll let me rephrase and
set of giving.

Two makes per kid with four drugs present
a body they give twice that dose for

makes for kids with only two drugs
parental bodies so the total number of

small molecule drugs are the same but
the number of antibodies it's loaded onto

is different so we have the same delivery
different penetration distance and

we looked into our hypothesis is
the same number of small molecules but

spread out more evenly in the tumor
will be more effective and

we took all the literature data over
the past ten years that we could find and

we analyze it and that's exactly
what we saw so in every case

these are tumor growth curve tumor size
in the vertical axis time on the X.

axis and so you can see you want low you
want a smaller tumor so it's more a.

Affective when you're down here all these
blue arrows here represent the fact that

when they spread out the spill over
more cells they have higher efficacy and

all these scenarios again the antibody
itself is not effective So

it's really heterogeneity of the drug
that's causing that efficacy Now there

was one exception we thought no we maybe
our theory is wrong it's incorrect but

if you look at this specific case this is
actually a very low potency payload so

in fact this is one scenario
where you need to concentrate

that drug on only a few cells so
you want a higher Dar in this case

in order to kill those cells otherwise
you diluted out too much and

it's not effective so it's all consistent
with this with this theory and

the idea from this is that we
really want to match the potency So

the I.C. fifty the amount of drug you
need to kill the cells with the K D which

corresponds to the delivery in vivo and
if you do that you have much more.

Effective drugs now I want to point
out that this is actually a very

counterintuitive the pharmaceutical
industry has been very concerned that

these naked antibodies are going to lower
efficacy that they're going to block

uptake of their drug and so there's been a
lot of concern with trying to avoid having

any sort of drug there's been some other

issues that we want to directly confirm
in a prospect of trial if this occurs so

we used the system the show earlier
to look at tumor growth and

advocacy and here are the results no
matter how you analyze it whether it's

tumor growth whether sponsor rates or
whether it's survival whenever you

increase the penetration you spread out
the drug more you have higher efficacy.

In black and yellow here these are showing
the untreated tumors and the high dose

antibody by itself there's no response
again these do not respond by mean

system or by inhibition of Perceptor to
the antibody when we just give the F.D.A.

approved drug at the clinical dose
we see of response which is good but

as we spike in a wonder why.

On a three to one or even an eight to
one ratio of the naked antibody we lower

tumor growth so eight times the amount of
drug again this is very counterintuitive

We're taking a drug that has no effect
by itself it actually blocks and

inhibits efficacy in vitro and when we
added in vivo we get a better response and

so it is really counterintuitive and it
sets up a scenario where actually having

a more potent drug counterintuitively can
be worse and you can be worse off because

if you increase the potency of a drug this
would never happen with a small molecule

but if you increase the potency of this
drug you're going to lower the maximum

maximum tolerated dose you're going to
having to give less antibody drug and

therefore going to penetrate
a shorter distance and

you going to have less at the Casie So
really it's about optimizing a matching

delivery with potency that
gives you the best effect.

I'm not going to go into too much detail
but we've also come up with some ways of

measuring single cell delivery and
efficacy so when we get into small

molecules in a second here we generate a
system where we screened near infrared for

force found some that are residual eyes
seem so you can put them on the antibody

when they're internalised they get trapped
within the cell others that are non

residual izing they diffuse away by
measuring these to near infrared channels

or you can come up with an exact amount
of payload number of molecules per cell

that are delivered intact and
those that are degraded and

we can follow this over time and
we can do this at a single cell level

with digestive tumors and we compare this
with pharmacodynamic markers to really

understand at the single cell level
exactly how these drugs are behaving and

so this is where a lot of the work with
some of our collaborators is going.

Now the final thing I want
to mention with A.T.C.

is before we move on to imaging is
bystander effects so one thing I

didn't mention is these payloads they bind
a micro to go there microtubule inhibitors

but for this one drug that I mentioned
there's no bystander effect so

by that I mean when it gets
internalized into cells the.

Molecule the effective or the active
agent cannot diffuse out of those cells

into adjacent cells but
there's a lot of interest if you look at

clinical tumors the target
expression is very heterogeneous So

there's many cells that don't express the
target so there's a lot of interest using

payloads that exhibit bystander effects
meaning they target one cell they can

diffuse out into an adjacent cell and
kill it so what we're interested in is

we've shown that this is very important
with without bystander effect but

maybe this will solve the problem maybe
we can just use bystander payloads and

that will allow deeper penetration and so
this is something we want to look at now

there's no effective way of imaging the
distribution of these payloads at least

not currently the doses that you have
to give are too low the sensitivity for

the resolution we need is not high enough
or with the radio labelled agents and

mass spec imaging is as not quite gotten
down to the resolution we need yet

hopefully we'll get there soon because
then we can test our our predictions but

until then this is really the only
way of understanding how these

results have an impact and I just want to
point out that even subtle changes this is

why some of these predictions are so
challenging even subtle changes in

the structure can have a major impact in
distribution this is just three different

hoaxed eyes many of you in the area
may be using these all the time but

there's only one difference in a fox
a group versus a hydroxyl group so

this changes the look of Felicity
the membrane permeable it easy and

if you have a molecule that's taken up
by cells very efficiently it binds to

D.N.A. and
you can see vessels in red D.N.A.

stain deliver intravenously
in a mouse tumor and

it only targets vessels right outside our
cells right outside the vessel whereas if

you lower membrane permeability the skin
get into cells you can use this version of

the host but it's much slower relative to
the fusion so this is an example where

the dumb color number in this case we're
talking about a membrane permeable it so

membrane uptake relative to diffusion is
is very low and so it can diffuse even.

He threw out the tissue and
you get much better distribution so

these are some of the facts we had to take
into account when looking at some of these

newer bystander results and from the
simulation what it turns out is the dumb

color number is very important again
this is in a mobilisation reaction so

diffusing into membranes and
it immediately binds you to the D.N.A.

or the microtubules so it's effectively
irreversible relative to the Fusion and

so from a theoretical standpoint if
you want a very efficient payload

you want to don color number around
one because if the look of Felicity or

the Permian billet is too high you'll
immediately enter an adjacent cell and

be concentrated on your
nearest neighbor and

never penetrate deep into the tissue and
so over time you can see it slows down and

this is a radius outside of the individual
vessel you can see it reaches fairly far

but not in a high enough
concentrations in the black and

red to kill on the other extreme if you're
too hydrophilic you don't enter cells

you can then diffuse evenly throughout
the tissue but you actually go back into

the blood before you ever enter a cell and
kill it so

you want the optimal level and so
this actually allows you to design

right now they're just using screening
approaches and then testing in cells and

in animals what payloads give the best
bystander effect but this allows you to

actually design what type of properties
you want on an effective payload and it

gives us some confidence that one of the
most common the P.D. molecules gives you

a dumb color number all right around the
optimal right around one to three and so

this is a good indication to us that of
these consistent with our predictions

I'm not going to go into too much detail I
just want to point out that we went into

other analyses to show that the summary
of the story is that bystander

effect can help so we used a lot of tumor
growth this is kind of the gold standard

data in the field bystander effect can
help penetrate deeper into the tumour so

it is better than payloads for.

BYSTANDER effects however you still
have better at the Can See if you spread

the antibody drug conjugate it out in the
tumor than you would from just bystander

effects alone and really the weight of
interpretation this is the bystander

mechanism of killing will never be as
efficient as direct cell targeting and

this applies not just eighty C's
book inception this applies to nano

particles as well so
let me just explain if you deliver

the payload right into a cell that's right
near the D.N.A. or microtubules to kill it

if you're you are relying on passive
diffusion here so if the drug

has to diffuse down a concentration
gradient into the interstitial and

then down another concentration gradient
into the second cell the concentration is

going to be much lower than it would
be in the initial cell and there's some

work in the nanoparticle field as well
some older work where they show targeting

of nanoparticles to tumor cells does not
change any of the uptake in that tumor or

the E.P. are affected nonspecific uptake
gives you the same amount in the tumor but

if you target it to cancer cell
you have higher efficacy and

some of the initial work was done I
believe with docs through Bisson and

docs servicing can diffuse through
three cell membranes and so

you can get into adjacent cells but
delivering them specifically to tumor

cells gives you more efficacy then
macrophage uptake and then diffusion into

other tissues and so there
are applications in particles as well.

All right that is just summarized this
portion of the talk again I want to tie

this in given a grandiose title you know
chemical engineers can solve anything so

we're solving these decade old problems
again you're all chemical engineers you're

mazing and I'm talking about our specific
application of this but the binding

side barrier problem this is been known
for decades the problem of this and

it's really been only in that maybe
the past ten or twenty years that

we understand enough to control it and now
we can show that it's a major problem but

there are multiple solutions so you could
use low affinity you could use low on rate

the reason we really like this spiking
in of additional antibody is that this

maximizes multiple mechanisms of action so
we're spreading our maximizing the A.T.C.

payload delivery but we're also maximizing
receptor signaling blockade if that's how

you're antibody works like with her too
and even more interesting to us and this

is where some of the work is headed is F C
effect or functions recruiting the immune

system because that's where a lot of your
durable responses are happening if you

can coat all the cells with those
antibodies you're maximizing payload

delivery and you're maximizing recruitment
of cells So in summary we've shown

that the penetration issue is critical for
these eighty season this is probably

impacting a lot of the drugs in clinical
trials currently in those in the pipeline

we can increase the penetration with
extra antibody with a carrier dose

a blocking dose and we've also shown
the bystander effects I don't want to.

Understate their importance they're
extremely important for killing answers

negative cells but you still want to
maximize antibody penetration but

for maximizing those bystander effects you
want to have the optimal physical chemical

properties and we can we can
design those based on structure.

All right in the last part of the talk
I want to talk about these two

imaging stories first one
is on type one diabetes.

So Type one diabetes as was previously
known as juvenile onset diabetes this is.

Millions of Americans and it's an auto
immune disease so your body incorrectly

recognizes your insulin secretion
cells as foreign It destroys them and

so your body has no way of regulating its
blood sugar the patients have to pick

their fingers multiple times a day inject
insulin for the rest of their life.

There's been a lot of interest in
the diabetes community in coming up

with an imaging agent for these cells so
a couple things I want to point out these

beta cells it's a little hard to
see here is a little washed out but

here's an islet they're located
in these islets of layering these

little small clusters we have
imaging agents so there's a extend

is a type two diabetes drug it binds
specifically to these cells and

this is a pancreas you can see all these
individual eyelets they pretty brightly

and so the idea is this when
a patient comes to to you with with.

Me with elevated blood sugar they've
often lost maybe ninety percent of

their beta cells so the disease has been
going on for a long period of time and

that's really the optimal time to
intervene However there's a lot of so

there's excess capacity so you can't just
measure insulin secretion ur blood sugar

response to know whether the diseases
started so that's why the community really

wants an imaging agent to see whether
you can detect a loss in the signal and

this project started back during my
postdoc where we were excited to show that

what we could do is using this extend and

as an imaging agent we could actually
detect the loss in beta cell mass prior

to the increase in blood sugar so in a
mouse model it was possible to detect and

then intervene of the early stage with
say immunomodulatory therapy to suppress

the immune system in order to stop
in potentially cure the disease.

So this is great there's lots of
groups working in this area but

it's been extremely challenging and

in fact a lot of the clinical trials that
have gone on with this imaging agent

have not succeeded and so we really want
to go back and look at the transport and

understand exactly how this molecule
is failing because it looks so

promising in animal models not
just biassed but many others.

And so in our lab what we did is we
started do a detailed investigation of all

the transport rates for this molecule and
I'm not going to go into a lot of

the details but we looked at coal at all
the nonspecific interactions we looked

at the viddy slow clearance no one had
ever quantified the down regulation over

time and one of it turns out one of
the most important things we did is

surprisingly although people have measured
expression with I mean his to chemistry

no one had ever quantified that specific
number of receptors Purcel and so

we did that in its fifty four thousand
that's reasonable about a million as high

one hundred thousand is moderate
ten thousand is low and

this is in a range that what we'd expect
but what was more interesting was

when we looked at our bio distribution
studies when we measured the amount

of specific probe that got
into the pancreas and we then

compared that to the number of receptors
per beta cell we could only account for

a third to half of the signal I'm not
going to go into the background but

it was specific signal we
could block it with peptide So

we knew it was the receptor but it was not
we couldn't it's a mass balance right it

can't be caused by internal as
a show of the beta cells so

this is a classic chemical engineering
we're closing the mass balance where is it

coming from and there's been
a lot of literature talking about

expression of this receptor in
other cells within the pancreas but

it was all controversy at
the end of bodies are terrible

they cross react with other molecules and
so there's been a lot of controversy and

not to mention this is a multibillion
dollar field there's a lot of concern if

there is expression on these other cells
because there's pancreatitis with these

drugs and so hopefully if drug development
standpoint there would be any expression

there but in our model it looked like
there was and so we wanted to analyze

this in more detail we used genetically
engineered mouse model a knock out of this

receptor and to just briefly go through
we can estimate the number of receptors

per cell it was very low it was less
than one thousand receptors Purcel which

is very difficult to measure we
could only do it because of a.

This very high affinity
specific peptide but

we can see a very high signal on our
eyelets when we use the knockout mouse or

block it is very low but
if we use the wild type mouse we do get

a higher signal and when we use the excuse
me when we use wild type we get a higher

signal knockout mouse or blocking it all
goes back to the same author forests so

this was direct proof that our molecules
specifically buying into this receptor

in the accident cells and now since then
there have been other groups that have

come out with using
multiple approaches of very

highly characterized antibodies
showing that this receptor is on X.

occurrence else of these are the rest
of the mass of the pancreas the not

the beta cells both in human and
multiple species.

And and so

this would really explain a lot of
the failure of this molecule in the clinic

even if you have a long term diabetic with
no beta cells there's still a lot of.

Receptor on the execution and so
just to clarify I forgot to point out very

important point earlier these beta cells
only make up about one percent of your

pancreas so what's confusing there is
a very few receptors per cell maybe one

hundred fold lower two orders of magnitude
lower on your off target cells but there

are one hundred times more prevalent and
so therefore they can dominate the signal.

What I like about our quantitative
approach is that not only does it

tell us of those receptors are there but
we know the exact number and

so this provides a potential solution so
I'm not going to go through all

the details of our modeling but we look
through and you can do an analysis and

really unlikely that we'll find
another molecule as good as extend and

you can do screen approaches look for
other targets on those cells but

in our opinion they're likely to fail
because you need of very high affinity

very low molecular weight interaction
that internalizes very quickly and

is expressed at a fairly high level and.

People been looking for that for
two decades and haven't found it so

the odds of finding it now we
think are not very likely so

how can we use this imperfect
agent to still get specificity.

So the simplest approach is blocking So
here's the problem we have these

non-target cells we inject a radio label
probe and it targets both cells we don't

have specificity if we get rid of the cell
in a diabetic we still have a lot of

signal so what if we inject the expression
level was one hundred fold lower fifty

fold lower What if we inject a low
amount of our peptide not tagged

to block those execute cells then
when we inject our imaging agent

there's orders of magnitude more receptor
there so we won't have blocked all of it

we don't specifically bind of those cells
and so that's what we wanted to do so

we went ahead and tried that and
it fails so it doesn't work at all and

no matter how many times the student
tried we would always get the same amount

of blocking so it's important to point
out radio labels are very sensitive so

we really need to block one hundred
percent of those receptors and

every time we increase the dose we
titrated up we block one hundred percent

of both the execution but
also our beta cells so

what's going on we don't have direct
proof but taking inspiration from

our chemical engineering if we look at
the size of these islets these are where

the beta cells are located there are fifty
to one to three hundred microns in size

if we look at the diffusion coefficients
for molecules in this size range

they're very rapid they can easily
diffuse throughout that tissue so

what happens when you inject and
you start increasing the dose to saturate

these cells outside of the dotted
line here they diffuse everywhere and

you're really limited by binding
it's a binding isothermal to K.D.

this is a thermodynamic limit that's
not what we want what do we want

we want a kinetically control delivery so
this is where we're

trying to get out of this thermodynamic
and just kinetically deliver.

Me deliver a small amount of
peptide into this region without

the fusing into the eyelets and
so rather than just.

Saturating everything in having one
hundred percent blocking everywhere

we decided to see if we could limit that
now the point of this slide is not to go

through all of it this is
there's two reasons for this.

I did a lot of great work he
just graduated from my lab so

if you run into him you can tell him
that I showed a lot of his work so

that's the main purpose of the slide but
the second reason I'm showing this

slide is because he did a lot of
great work on peptide engineering

I'm not going to time to talk about
it today but multiverse is ation.

We used by orthogonal chemistry to
stabilize the alpha helix increased

binding affinity produce stability etc and
we had a lot the main

point of the slide is we had a lot of
tools in our toolbox can we use any of

these to slow down diffusion and so
what we chose was a lipophilic modified

peptide to hopefully bind a plasma
proteins slow down to fusion and

get us into this kinetically controlled
blocking and it turns out everything

we've done so far it works we haven't done
this in diabetic mice yet but we do when

we do this in healthy mice we're able to
specifically block basically one hundred

percent this is within the noise year of
the accident pancreas Now notice we're

blocking about eighty five percent of the
end of that might sound like bad news and

it was certainly higher than we had
hoped but we're never going to be able

to completely block the fusion in
the islands what I want to point out is

because the expression is so much higher
We have almost ten thousand receptors for

so which is many fold higher than
what we actually need to detect with

the radio label to Agent So
this is still within.

Our parameter space where we can detect
these islets but now detect these beta

cells but now we don't have to worry about
specificity because we blocked everything

in the non-target tissue so that's where
we stand now with this project but we.

Now have a translatable path
forward that we're testing in

diabetes models of mice in order to
get a specific imaging probe for

detecting this type one diabetes so it's
really twofold using the chemical injury

mask balance to detect what the problem
was and then look at the kinetics

in the transport to go from a binding
limited regime to a delivery in

this case permeability limited regime to
specifically block the execution pancreas

are and the last thing I want to talk
about in the remaining couple minutes here

is a screen approach so now there were an
attorney in the corner here we have these

simulations we can design molecules
with different properties we want to

do something that no one has ever done
before so we selected these problems that.

Are very important for
detection and scream so

if you look at the model feel a lot of
the results coming out of the past year to

looking at the epidemiological scale have
shown that there's been a massive amount

of over diagnosis we're detecting tumors
earlier we're detecting them small but

those patients likely would have never
died from those tumors we're not doing

is unlike colon cancer and polyps screen
we're not seeing a large decrease

in deaths or later stage tumors which
indicates that a lot of those tumors would

never cause any problems in the patient
in fact estimates are that this is maybe

a four billion dollars problem the amount
of over diagnosis and overtreatment of

breast cancer so how are we going to solve
this well one of the major problems is

fear M.R.I. or ultrasound or
a lot of these imaging is still

focused on anatomy or maybe physiology
if you inject a contrast agent

even though therapy has many years ago
gone to a molecular level treatment so

we need to get molecular information
to really identify which tumors

are malignant we need to remove in which
we can just do watchful waiting so

we need molecular and spatial information
the blood tests aren't specific enough and

molecular imaging can do this the problem
with molecular imaging is it typically

uses radio label to agents that
are costly and involve radiation so

cancer risk so in fact there was actually
a clinical trial where they used

radioactive sugar in Japan and they they
screened a large portion of the population

it was very controversial both from a cost
standpoint and giving out these doses of

radiation to potentially healthy people
and they detected some tumors earlier but.

It's not really a significant
benefit we need a better way so

our approach to novel approach is let's
do generate a disease screening pill so

a molecule with a near friend floor for
the can be taken orally absorbed into

the intestinal tract target sights of
disease and then be imaged non-invasively

with near infrared light so the two
applications we're going after are breast

cancer and women might be short on
time but also rheumatoid arthritis and

so why are we doing this well delivery as
opposed to I.V. is really important for

safety so even though and if lactic shock
is life threatening it's very small from

I.V. delivery but there are forty million
mammograms a year and so even point one

percent is too high when you're dealing
with potentially healthy individuals so

safety is important issue cost you don't
have to deal with sterile delivery

intervene Asli you don't want to deal with
health care professionals giving an I.V.

and so cost becomes very important as
well if you want to scale this and

finally patient compliance so
patients strongly prefer you know nine to

one either be or
delivery over interview just injection and

not only that but
all of these agents take hours to a day or

so to target so if you have to deliver
I.V. you come in you get your I.V.

injection you're monitored to
make sure you don't go into shock

you leave come back later you stay there
for hours it's just not going to work for

patient compliance or delivery
they can take the pill at home and

come in and get scanned so
if this is so great why is no one ever

done this before well there are multiple
arguments it could be made there but

I think fundamentally one of the main
ones is that the properties you need for

delivery are the exact
opposite of what you need for

high targeting efficiency so
you need small lipophilic agents for

absorption the G.I. track you need
large hydrophilic agents for specific.

It by need to targets also this is one
of those scenarios where you do a lot of

simulations but the one parameter we
didn't know was the or absorption of these

molecules they're outside of the Lipinski
rules for those of you have heard of that

and so there's not a lot of research
in the pharma industry on this so

we generated a series of agents we had
a lot of design criteria that we used to

make sure that we could come up
with something successful and

then we went ahead and
tested all these and

to jump ahead to the punch line it works
really well and so we're using a or

the topic model of breast cancer here and
we deliver this probe by organ Vosh we

can see after six hours there's actually
a fairly high signal in the tumors but

it's it hasn't cleared for the background
there's a massive amount in the gut and

then over time it's retained in the tumor
it clears from the rest of the tissue and

you get contrast levels that
are sufficient to detect tumors

we estimate from clinical trials less
than two centimeters in size and

clinical Depp's so I did point out there's
been a lot of work in the near Fred

tomography world looking at this but
they're looking at just plain dies so

they never took off because you don't
have great resolution you're not binding

to anything specifically serially not
any better off than say an M.R.I. but

here we're getting molecular information
from the target we picked and if we use

a non-binding stereo isomer we don't see
any signal in the tumor cells behaving as

expected and really what we found out from
a molecular engineering standpoint is that

the negative charge is the key so it's
very counterintuitive we have much higher

absorption the what we expected certainly
not at the levels of therapeutics but

in an imaging agent you can
tolerate a lot more variability and

it's only given once you can
tolerate higher doses and

they're still reasonable about five mix
but what we found is these molecules

are highly negatively charged but if you
look through the literature a lot of that

negative charge there is evidence
of opening of type junctions and

increasing the diffusion through cells and
so this is what we believe

is going on as we increase the neg.

Charge we had higher absorption higher
excretion in the urine and not only does

the negative charge help with absorption
it reduces nonspecific sticking to

proteins it actually slows down clearance
is not great for an I.V. agent but

based on a simple one compartment model
if you have the same absorption rate and

you slow clearance you get higher
concentrations so it's important for

targeting there and we've been a lot of
work looking at residual zation of dyes

and shown that this traps it within those
cells and so that allows us to wait days

in order to see that signal so
I'm going to cut the last part out.

For time so we have a chance for
questions but I want to point

out when we looked at the cells we chose
this target for high sensitivity it's.

Alpha Beta three binder it's
a pep to Dometic And so

it's present on blood vessels tumour
associate macrophages and cancer cells and

that gives us a very robust signal we're
looking at now is can we pair this

with a second probe to help with our
specificity so we have two different

wavelengths that we can interrogate to
give us a very high sensitivity and

specificity to the tumour there's a lot
of interest but for the sake of time I'll

skip over this in doing the same thing for
a room with her through this because

there's been a great transformation
in our a way where we've gone from

giving pain killers to actually stopping
the disease and there was a recent

analysis of a clinical trial where you
can actually cure these patients if you

can detect them early enough and just give
them a one year treatment a method track

straight to suppress the immune
system you can take them off their P.

and L.
never get the disease but

there is no way of identifying at
early stages which patients have

rheumatoid arthritis and so that's why
we're developing these imaging agents and

they work in mice as well so I won't go
over the details and I want to point out

who did some excellent work as
an undergraduate my lab who is now here at

Georgia Tech but she built a console
model that we can use infrared light.

Simulations to show how
the levels of targeting we're

getting would translate
to clinical Depp's.

So I just want to point out the vision
on the imaging side there's been a lot

of work this is an F.D.A.

approved agent for R.A.
this is these were some of the tomography.

Instruments used for breast cancer but
these were with nonspecific probes so

really this is what's limiting this
field is identifying probes that can be

easily deliver and give you specific
results so in summary we've showed that

the heterogeneous delivery of antibodies
is causing major problems with efficacy

but we can design better agents using
our simulations quantitative imaging of

uptake for type one diabetes was used
to identify what the problem was and

identify a solution to help block
some of this off target uptake and

we can use these simulations
design molecules for

a disease screening pill that could
take by mouth and then read light for

a needle free and radiation free disease
screen approach but to summarize I

think the chemical engineering
principles the quantitative P.T.E.

simulations are using before we go
into animals paired with those in vivo

experiments provides a really powerful
tool to develop new imaging agents

as well as better therapies which we
talked about with eighty cities so

thank you very much for your attention
I'd be happy to answer any questions.

And then.

I.

Read it's a great question
the question was.

I'm going to end it and a particle is
either that that small example of some of

the targeting you know what space to do we
think particles will play a large where so

that's a great question
I think run of the.

Big gaps in our knowledge for

predictions is in some of these in
this non the Penske space of molecules

that are five hundred to five thousand
battens in size every move in rather But

how do you get molecules a don't normally
diffuse across membranes into cells and

I think those are a lot of
opportunity in that space for

both the fundamental understanding of how
these molecules get into cells you know

how do we simulate a staple peptide how
do we simulate nanoparticle delivery so

that membrane permeable really question
is still one of the large unanswered

pieces in the model also I think there's
a great opportunity there to engineer

materials for intercellular delivery
the other that I'm interested in

where I think you know part of us
have a great opportunity is with.

G.
and

engineering the immune system because
within a particles you really have a great

opportunity to engineer both the video
the number of different types of receptors

the types of peptides you're delivering
the interaction with different cells in

the immune system so that's another area
where I think there's a lot of opportunity

for nanoparticle work to really contribute
and there's some exciting work in immuno

oncology that I think any
particles can really address.

That is a great question so

we have talked with them and
I I think they should do it I think

it's something that they're
they're not focused on now so

we've talked with them several times and
they they they really should do it they

especially I think what what I
think they should invest in is more

with the combination therapy with their
immuno oncology agents so they're P.D.-L

one inhibitor that blocks
the suppression of the immune system so

basically activates the immune system
there they're combining that currently in

clinical trials with cat Sila and with
receptive what I think is that three way

combination to both deliver the cat Sila
because that the payload can stimulate

the immune system but also coat all
the cells with the antibody and

then allow some suppress blocking of
the block suppression of the immune system

I think that would be a very synergistic
combination right now I think

they idea that blocking uptake is
going to help is still very foreign

to them I think they're still worried
about blocking uptake hurting efficacy and

you know we haven't published work yet
and no one has really published

any work showing this yet so I think
there's still some skepticism there but

some of the other pharmaceutical companies
were working with they haven't published

their data either but they see the same
effects where the more potent drug in

vitro is less efficacious in vivo and
so I think where really this

will initially enter the market is more in
the current design of the next generation

eighty Cs I want to point out one of
the drugs that has fast track status that

in solid tumors is with that lower potency
payload and they've seen good responses

in Phase two clinical trials there
giving twenty makes particular antibody

versus cats which is three point six So

we believe there are saturating
much more of the tumor and

we believe that's one of the reasons why
that one is as been so successful so far.

Right great question so
the question was on diffusion go fish and

binding affinities said it right so
great question for for

a lot of the molecules we have correlation
so of course they have to be measured

experience experimentally originally
we can make reasonable estimates based

on the molecular weight let's
say of a hydrophilic protein and

then with the small molecules we can make
reasonable estimates based on lip of

Fla city hydrogen bonds etc but I like to
think of it as a layered approach we're

never going to be exact same with mining
affinity you can do docking studies things

like that but you're never going to be as
good as experiment were allows you to do

is make an initial simulation then when
you get some in vitro data it's a little

bit more effort but you get that in vitro
data and you can refine the model and

make even more accurate predictions and
then you can go into the animal models and

see if those predictions match with your
experiment if there is say interaction

with the protein that you weren't aware
of then you can add that to the model so

you can keep building in complexity but
I think it's important to at least be

able to put an initial stake in the ground
saying based on these properties this

is what we expect because I can identify
if you have very unexpected behavior and

it can help you narrow down
where that's coming from.

Some of the experiments we've done we've
done I mean you can do FRAPS a photo

studies into vital microscopy in tumors
in an animal so that's where some of

the studies come from you can do three
dimensional tissue culture as well so

you've done a sphere it works to look
at some of the diffusion there so

those are two of the primary ones they're
in for diffusion coefficients that's one

area where there has been a decent amount
of work by pharma companies and many

others so we can look at some of those
datasets to help fix our correlations.

Thank you.

Thank you.