2021 NIH Synthetic Biology Consortium Meeting – Day Two

-Good morning. I want to welcome those of you
who were with us yesterday and give a nice welcome
to those of you who might first
be joining us today. My name is Michelle Berny-Lang,
and I'm a program director from the National Cancer
Institute. My background is in engineering, so I've been a long-term member
of the engineering community but relatively new to
the synthetic biology community, and I've really been struck
by what a collegial group it is. I really see so much
sharing of resources of tools
and collaborations, and I really think
it's fantastic that there's that recognition
that to continue to advance
this field forward, it will really need the efforts
of so many and the space. So thank you for letting me
be a part of it. I'm really encouraged
by your efforts and your ways
to work together, and I also know you probably
have about five other occurring meetings
on your calendar right now.

So thank you for taking the time
to be with us today and be part of this community. I wanted to start by quickly
recapping what we covered yesterday. When I had a chance to look
at the numbers during one of
the breakout sessions, I saw that in one
of the breakout rooms, we had close
to 200 participants. So I think yesterday
was fantastic for attendance, and I think even more
than having individuals tune in, we had great engagement
with scientific questions and then the interactions
with our NIH community. So we kicked off yesterday
with the scientific highlights from the synthetic biology
and cancer program where you got to see
really kind of a plan for what these investigators
will do over the next 5 years, and if you weren't
able to attend and that piques your interest, I definitely encourage
you to go back and check out the reporting.

We really want to deepen
the connections between the broader
synthetic biology community. So if you have ideas
for collaboration or even might want to find
a more formal way to participate through
the affiliate membership process that we'll be establishing,
please reach out to me and we can help to facilitate
some of those connections. At the same time
as that session, we had the grant writing seminar for our early career
investigators, and I hope you found
that informative and hope
that you'll be next up submitting your applications
to the NIH, and then I think it was
really exciting to see how many different
components of NIH, to have 10 different components
in the same room that are demonstrating
their interest and their ways that they want
to support synthetic biology.

So I think each of the groups
that joined in is already supporting this space and had a great set
of opportunities and ideas for how to bring in more of your
research to the NIH community. So I hope that you really leave
with that message that NIH is very invested
in synthetic biology, and there are opportunities across a huge range
of biological applications. So think as Dave was mentioning
yesterday, I hope you started to see
some connections for how your research
might fit with NIH and some people
that you can work with, some program directors, as you work to develop
your applications.

Often as a program director, I find that investigators
are hesitant to reach out to me and are concerned
they'll be a burden, but it's really my job and we're
here to foster your research. We really want to be
deeply connected to the research community and help to make your research,
as it fits, be supported by our institutes. So please continue to engage.

I hope you had a chance
to meet some faces yesterday and continue
those conversations. Even just seeing your latest
papers or your latest research, it's really valuable for us
to keep on what's going on
with the scientific community, and then if you continue
to develop that relationship, we can help you out
with letting us know what other new and
upcoming opportunities exist. So I think it's really nice
just to continue to develop those connections between
the research community and NIH, and again, that's really what we
want to do as program directors. So please definitely
reach out to us, but I think what's
so exciting right now, we had fantastic attendance
at this meeting, just a really well-developed
synthetic biology community at the same point that we
have strong interest at NIH.

So I think we're at a really
good opportunity to continue to push those approaches
and to have them be more and more a part
of what NIH is supporting. So thank you for helping us
push through that direction, and I think in terms
of program directors, while you saw a handful
of faces yesterday, I can definitely speak
for the NCI. You saw three representatives, but behind us are so many
other colleagues with experience and interest
in this space. So there are a broad number
of people that are really here to help and to continue to foster
synthetic biology at NIH. So moving on to today,
we're going to have our day bookended by some focus
on translation.

We'll start with investigators
who have moved their approaches through the FDA regulatory
processes, and then we'll end the day
with representatives from FDA before Dave
gives us the final sign-off, and then in between that,
you'll get to see some of the breaking
synthetic biology research from these synthetic
biology showcase sessions. So I hope that you come to those
and really get a chance to learn about what's going on
in this space, and I hope this inspires
some potential collaborations and some
additional investigators that you may be able
to work with. So we'll have those two
synthetic biology showcases and then a Q and A panel
after each of those. Today is a little bit different
than yesterday in that everything
will occur in this room. So you're here now,
and you're in the right space and can stick with us
throughout the day, but it's just really encouraging
to see your enthusiasm, to see the enthusiasm
across NIH, and really excited to see
where we go in synthetic biology and to see how much
of a bigger part it's poised to become at
the research that NIH supports.

So thank you so much
for being part of this meeting. I don't have 15 minutes
of content to fill so what I think I can do is give some more time
to our next session coming up. So I'll just go ahead and
introduce our next session and our next speakers. So I'm pleased that I get
to kick off the day, and as I mentioned this
is a session on investigators and researchers who have experienced
the transition of their techniques
and approaches that have gone through
the FDA regulatory process. So we'll get to hear from
Doctors Tim Lew and Gary Lee and they'll give a presentation on designing intelligent cell
and gene therapies. Before I pass it off to them, I just wanted to briefly
introduce them both.

Dr. Tim Lu is an associate
professor of biological engineering, electrical engineering
and computer science at MIT, and he has a dual hat
serving as a cofounder and chief executive officer
at Senti Biosciences. He's trained in engineering
and holds an MD
from Harvard Medical School and a PhD from a Harvard MIT
medical engineering and medical physics program, and for his synthetic
biology research, it's focused on the design and
application of synthetic biology that really spans
broad biological contexts from nanotechnology, amyloid associated diseases
as well as infectious diseases. He'll be joined in his
presentation by Dr. Gary Lee who serves as the chief
scientific officer at Senti Biosciences. He trained in chemical
engineering and earned his PhD
from UC Berkeley, and for more than a decade
has been leading cell and gene therapy programs
for human applications. So thank you for joining us. Thank you for starting off
the day, and we're pleased we get
to learn from you. -Great. Well thank you so much
for the time. Really appreciate
the opportunity to be here. So I'm Tim, and I'm going to
give an overview of what we're doing here
at Senti Bio, and then about halfway
through Gary will take over.

He's really the translational
expert in the room and has been really
through multiple rounds of taking programs forward
into the clinic. So please feel free to ask
questions along the way, and we're happy to share
what we're up to. So as mentioned, you know, both Gary and I
are working at Senti Bio and really the focus
of this company is to try to translate
synthetic biology technologies that we're all familiar with
and advance these forward for cell
and gene therapy applications. We're really excited all
the progress that's been made in the field
in the past several decades.

For me, as a faculty member
over the last decade or so, really a lot of the synthetic
biology techniques that have been developed
and translated and now possible
to do in mammalian cells has been quite exciting and at the same time sort of
the rise of AB gene therapy, NK cell therapy, T cell therapy
et cetera in the clinic has really opened the doors
for some of these technologies to try to make them out
into the actual real world. So at Senti we've pulled
together many of the sort of gene
circuit componentries that have been developed
previously in academic labs and have really tried to combine
these together to allow us to program
a wide variety of cell types and gene therapy vehicles. These include, you know,
immune cells, include stem cell therapies, include gene
therapies et cetera. We believe there's an
opportunity here to really improve
the profile of these cell and gene therapy products
in terms of enhanced precision, efficacy and control as well,
and so what we've done at Senti is really trying
to make this a platform, both for internal applications
as well as for partners.

Internally, we'll share some
of our ongoing efforts around allogeneic CAR NK cells
for oncology as well as collaborations
that we've established now with several
large pharma players, including with the Spark
division of Roche in the AB gene therapy space as well as the BlueRock
division of Bayer pharma. So I started this company
together with a couple of great folks
including Philip Lee who is currently
the chief technical officer and lead of all
the manufacturing efforts. One of the areas that in
the cell and gene therapy space, we've invested quite
a lot of effort into is in process development
and manufacturing. As a, you know, primarily
synthetic biologist originally by trade, really I didn't necessarily
appreciate the complexities of manufacturing
these products but I think now
within the company, I think there's a significant
emphasis on being able to, you know, scale up manufacturing
that certainly impacts, you know,
the accessibility and the costs of these products for patients, and Philip has
really been spearheading a lot of those efforts.

Jim Collins as well as Wilson
Wong, our scientific cofounders, obviously well known
to this community and pioneers in this space
of developing gene circuits for mammalian
therapeutic applications. So why do we really take
the impetus here to start Senti, and it's really centered
around some key challenges we see with current drugs, regardless of whether you're
talking about CAR T, CAR NK cells,
to gene therapies et cetera and these really
are centered the idea that current products
are not really able to match the complexity
of disease in a very direct way, and to really dive
into a bit more detail, one of the challenges
currently with these products is oftentimes
they're basically designed to go after single target, but that single target
may not be clean enough to really distinguish between
disease and healthy cells due to target heterogeneity
or disease heterogeneity, and as a result
you're always trying to thread the therapeutic window of
trying to hit the disease cells while sparing
those healthy cells.

That's number one,
a major challenge. Number two, is oftentimes the
diseases that we're going after, if you think
about autoimmunity, if you think
about cancer or others, these diseases are complex
and can invade, you know, single
target drugs, you know, if you're
going after one target, you know, the disease can evolve
and escape from that threat, and so that's another challenge
in terms of existing molecules going after
complex diseases. Number three is a narrow
therapeutic window. I think we all know that there's
some safety challenges that have been observed
with for example, CAR T cell therapies, some recent challenges
with the AB gene therapy side. Much of this comes
from the fact that we can't actually regulate
these drugs very well, these cell and gene
therapies very well after we deliver them to patients, and so if we are
going after diseases with a narrow
therapeutic window, you know, the cell sort of acts
the way it acts, and we really don't have
good control over that.

Number four, dynamic
disease conditions, oftentimes many
of the diseases that we're going after
actually only occur in certain parts of the body. For example, in the brain
or, you know, in the muscle, and you want to be able
to hit those tissues without hitting everything else, or you want to be able
to have a therapy that actually changes
itself over time, and that's difficult to do
with current static medicines.

So we think that it's possible
to use synthetic biology to overcome
these key challenges. Obviously you guys all know
that, you know, these genetic circuits
that, you know, the field
has been focused on building, you know, oftentimes are
inspired by natural systems, but we've now put them
together into artificial ones. There's really four key
categories of gene circuits that Senti is focused on
because we believe it addresses a wide range
of therapeutic applications including logic gates,
we call multiarmy, the regulator dial
and smart sensors. One of the things
that we wanted to do when we started the company
is to ensure that we had a powerful platform
for designing and optimizing
these gene circuits. Instead of just simply trying
to build these things one by one which, you know, certainly
we've done in the past, and we can get it to work, but we wanted to really make
this an industrialized process. So this includes, you know, a lot of computational and high
throughput design techniques. Starting on the left-hand side
with the core gene circuit, synthetic biology design team where we've developed
a variety of algorithms that help us to number one,
identify targets we want to go after and two,
really serve as a design team for putting together
sort of initial, you know,
gene circuit type designs.

Then there's the build side. This includes, you know, DNA
engineering, vector engineering, so, you know, we've gotten
really good at both lentiviral as well as retroviral as well
other vector components. So basically being able
to manufacture those in-house at scale,
as well as cell engineering so being able to take
those vectors and transduce a variety
of different cell types internally
has been important because we want to be able to
generate libraries of the work that we're going after. On the test side, you know,
at Senti internally we initially focused
on oncology applications. So all of the assays that we
built up are really centered around testing the genetic
circuits in that context. So that includes in vitro cancer
immunology teams who can help us characterize
the activity of, you know, not just
the immune cells themselves but also their interaction with
other immune cells as well as, you know, the ability
to target cancer cells. We also have an in-house
vivarium where we have — We can rapidly establish
those animal models there and do testing,
and that's all important because as everyone knows here
that, you know, this design, build, test, learn cycle
is quite important in our field.

So we've gone through
the crank multiple times here. We have generated what we call
our central knowledge database. Basically it's a library
of gene circuits that we know that work well
versus the ones that don't, and that certainly allows us
to really accelerate design every single time we go
through this design cycle. So in order to really try
and advance these programs forward
we've centered a lot of our efforts
in those four categories, those fundamental disease
challenges I mentioned earlier. So number one to try to address
target heterogeneity, we've been working
on logic gates allowing us to basically
target multiple antigens rather than just a single one,
and once we can do that, we're able to detect
signatures of disease rather than just
a single target, and by doing so,
we think we can, you know,
enable much greater precision into a variety of cell
and gene therapies.

Number two to go after
disease evasion challenges we've, you know, been optimizing
multiarmy gene circuit so we can basically
have a single cell that expresses multiple
payloads at a time. Obviously this sounds
a little bit trivial, you know, we all know how
to express multiple genes, but actually being able
to express, number one the sort of
right combination of genes to be able to mix
and match rapidly different payloads
and then characterize the activity
has been a focus of the company. Two is making sure we can express
those at the appropriate levels. Oftentimes we just want to
maximize expression of both, and so being able to,
you know, again optimize the position of these genes
in the construct, you know, the way we express them in
terms of multicistronic nature, et cetera is important
to try to drive, you know, the maximum efficacy.

Number three, I mentioned the lack of sort of control
with these products in terms of going
after narrow therapeutic window. We've developed a variety
of what we call regulator dial gene circuits
where we can use FDA approved oral drugs, ones that have good
PK/PD properties to titrate up and titrate down the activity
of these sort of products, and then lastly, you know, the idea
of addressing dynamic diseases.

You know, basically
we can build sensors now that respond to a variety
of disease biomarkers. So, you know, we're not just
talking about cell surface antigens
which, you know, we may all be familiar with
the CAR T, CAR NK cell where we have a receptor
on the outside that recognizes a disease antigen,
but we can also build sensors that sense soluble molecules,
you know, if you have a soluble cytokine
or soluble factor and we want to titrate
the activity of the product, we can build receptors
to do that, as well as sensors
inside of the cell.

For example, synthetic promoters
that respond to certain transcriptional
signals inside of the cell. So the way we've done this
is now we've built up a respository
of internal gene circuits including, you know,
in the logic gate category, or gates and not gates,
and Gary will tell you a little bit about that later
in terms of application for oncology
applications for those. On the multiarmy side, we've built
a variety of strategies for expressing
therapeutic payloads including different combinations
and screening through those, as well as a technology
we called calibrated releasing and Gary is going to share more
about that later in terms of being able to tune
the amount of a payload in terms of
how much is on the cell surface versus how much is secreted. In terms of the regulatory dial,
we have different classes of small molecule
regulated promoters that can titrate up and down
with FDA approved drugs, and then again there's different
classes of these smart sensors.

Internally, we very much focus
on promoter design that can be targeted to certain
cell types or certain cell states. One of the key things for us is
to showcase that we can actually do this across
many different types of cells. Internally, we're focused on
building a pipeline of NK cells, but we've partnered
in other areas as well, so clearly trying to build this
as a platform company similar to,
you know, genome editing, it's quite readily possible
to deploy these in many cell types
or viral vectors. So let me give you
a couple specific examples of the types of technologies
we've been building at Senti and some of the justification
for why we're doing that. So one of the key classes
of gene circuits that we're quite excited about
is the logic gate. Everyone here is very familiar
with what a logic gate is and those that — There's different flavors of
logic gates that one can build.

Right? There's or gates, and gates,
not gates et cetera. At Senti we're very much focused
on the not gate for a couple of reasons. We want to try to improve the ability of our therapies
to target cancer cells and avoid
killing healthy tissues, and the reason for that
is we know that for many tumor antigens
that one would like to go after, those tumor antigens are
not perfectly cleanly expressed. Oftentimes those tumor antigens are not just expressed
on cancer cells, but they're also expressed
on healthy cells.

In some cases,
we're okay with that. So for example,
if we think about CD19 or BCMA, even though those can be
expressed on healthy cells in addition to cancer cells,
it's sort of okay that we go in and kill everything
that expresses CD19 because the patient
can still survivor in those sort of contexts. However, in many solid tumors as well as other liquid
tumor indications that's a problem where
if that healthy antigen — If that tumor antigen is also
expressed in some healthy cell, you're going to generate
pretty significant toxicities against the patient
and that limits ultimately the therapeutic efficacy as well
as safety for that patient.

So there's a couple ways
to try to address that problem using logic gates. The fundamental idea is instead
of recognizing a single target, what if we could recognize
multiple targets? So that's really what we're
trying to illustrate here on this particular slide. On the left-hand side is a
picture of the engineered cell. Here we're showing NK cells
as an example. The way we try to address
that problem is by engineering the not gate. The not gate is composed
of an inhibitory CAR receptor which is shown here in purple. So this inhibitory receptor
is able to block the activity of
the conventional activating CAR, and so there's two scenarios in which you can see
this functioning.

So on the left-hand side, let's say
this recognizes a cancer cell that primarily expresses
the tumor-associated antigen, well that activating CAR which
is in green should recognize that antigen for
killing the bad cancer cells. That's pretty straightforward. Now what happens if that antigen is also expressed
on a health cell which is on the
right-hand side of the green. So if that happens, a conventional CAR T cell
or CAR NK cell would also kill
that healthy cell. However, in this case
what we've done is discover what we call
a safety antigen that's highly expressed
on the healthy cell and not expressed
on the cancer cell which is shown here
in this purple pair here, and what this purple
antigen allows — Interacts basically
with the inhibitory CAR and the inhibitory CAR
is basically designed to block
the activating CAR and basically spare the
healthy cell from being killed.

So this basically implements
the boolean logic do not kill healthy cells
which gives us this ability, we think, to rescue
a significant portion of the tumor-associated
antigen space and protect the healthy cell while still maintaining
killing of cancer cells. Now the alternative to doing
this is the and gate which we've seen many examples
of in terms of being able to recognize
two tumor associated antigens and only killing cells
that express both. We personally prefer
the not gate to the and gate in that we believe
it's less susceptible to tumor escape considerations in that,
you know, with the and gate if you lose antigen one
or antigen two then the cancer cells
can escape. The second thing to point
out here is this mechanism
of the not gate is actually really done
at protein signaling level. So the way this functions
is really a lot of engineering around the ITIM domains
of the inhibitory receptor and therefore we've shown
that this can operate on a very fast time scale compared to relying
on transcriptional logic. We think that's important if we
want to be able to distinguish between healthy and cancer cells
on a fast time scale.

So we've now tested
and optimized this in a variety
of different settings. Again, Gary, will share with you
that a little bit later in all contexts, but this is just one example
in vivo context where we injected healthy
human cells in green, and then human cancer cells
in yellow into a mouse along with the CAR NK cells
that either contain the not gate or do not contain the not gate. So the middle figure here,
number one, we simply looked at overall killing
of the number of human cells, and we can see that
in both cases with the CAR NK cells
with or without the not gate, you do see
a significant reduction in the total number
of human cells, however, there's a bit of sparing of
those cells with the not gate, and so what we wanted to do
is dig in further to look at what cells are actually
being spared in that population, and on the right-hand side
if you actually look at the percent
of healthy cells overall, so the percent of healthy cells
out of all healthy cells plus cancer cells,
we see that with the not gate there's a significant enrichment of the healthy cells
in the mice, and what that means
is essentially that we're still able
to kill the cancer cells, but we are sparing a substantial
increased protection of the healthy cells
using the not gate.

So we're pretty excited
about this platform, and again we'll show you later in terms of the types
of cancer application we're trying to go
after with this. Number two I mentioned
the multiarmy gene circuits. I mentioned again mixing
and matching different payloads. One of the areas that we're
interested in especially in solid tumors,
the arm, our CAR NK products to really make them more active
in the tumor microenvironment which can be highly sort of
suppressive for immunotherapies.

We also want to be able
to stimulate surrounding cells. So basically not just sticking
the CAR NK themselves but also recruiting other
aspects of the immune system. So we've built a library
of different immune payloads. This is just a small
sampling of them. You know, we have about,
you know, I think 30 or 40 of these
that we've prioritized, and then what we've done
is mixed and matched a lot of these
together, so built, you know, many, many different pair
rise combinations, and even triple combinations
that we can express at high levels
within the cell therapy context, and then what we can do
is characterize these both in vitro
as well as in vivo models to determine
their ultimate efficacy. So as an example, there's two
particular cytokines that are quite unique
here in this example that we've engineered
into the CAR NK context which is important in potentiating
their overall killing activity. As many of you folks know here,
with CAR T or CAR NK cells, you know,
sort of repeated stimulation and exposure to cancer cells can exhaust
those cells over time.

So one of the things we want to
try to do is reduce that exhaustion and maintain sort of maximal
killing efficacy. These two particular cytokines
in combination with each other, which is shown
in the light blue, do allow us to robustly
kill tumor cell populations compared to CAR NKs
that express no cytokines which is in the dark blue,
as well as CAR NK cells that express individual
cytokines on their own. So this is one example of where
this sort of ability to go after multiple sort of stimulatory pathways
at the same time can lead to significant benefits
for the overall product profile. Number three is the
regulator dial. Some of these payloads that
we're interested in expressing, you know, can really benefit
from in vivo control. As an example, if you think
about IL12 as a cytokine, you know, many people believe
that's a quite potent cytokine based on a lot of
translational research that's been done
over the last 2 decades.

IL12 has also been tested
in humans both as a, you know, freely delivered cytokine as well as engineered
into T cell therapies, and what people have seen
is that IL12, you know, has a somewhat narrow
therapeutic window, in that if you have too
high levels of it systemically and aren't able to control
how much is there, you can quickly run into, you know, significant
adverse events in patients without really generating
the level of significant immune response that you're interested
in looking for. So one of the ways we believe we can overcome
that sort of challenge is really enabling in
vivo control of these products using exogenous
FDA approved oral drugs. Now this is not a necessarily
a new concept.

There's been many examples
of these in the past both in the academic literature
as well as in certain clinical studies, but one of the key challenges
we've seen previously is that number one,
many of the small molecule drugs that are being used
to control cell or gene therapies in vivo have not ideal
or poor PK/PD properties, so, you know, they may have
a very short half-life, they may not — May require you to really
constantly dose the patient which could be
a significant challenge from a clinical perspective.

Number two, the on, off ratios
that you can achieve with a lot of conventional
technologies may be quite small, you know,
less than an order of magnitude. So really our focus here
was twofold. One is to be able
to build drug switches that can rely on,
you know, drugs that have actually
good PK/PD behaviors in humans. So these include FDA approved protease inhibitors
for hepatitis C virus. It includes IMI drugs. It includes tamoxifen,
as well as others, and the reason why we have
multiple classes of these is depending on which sort
of tissues you want to target or which patient population
you want to target you may actually choose
one versus the other. The second thing we wanted to do is build multiple flavors
of the regulator dial.

So in some cases we want
to build an on switch where adding
the drug turns it on. We also wanted to build
an off switch where adding the drug
turns something off, and we also wanted
to really maximize the on, off ratio
of these sort of products. So here is one example of the
regulator dial that's now been engineered
to control IL12 secretion in response to a molecule
called grazoprevir. Grazoprevir is a hepatitis C
virus protease inhibitor as I mentioned earlier, and what we wanted to do here
is really try to get, you know, a couple orders
of magnitude of control of IL12 secretion in vivo. So what we're showing you
on the left-hand side is the differential in IL12
that can be generated in the presence
of grazoprevir on day 4. We can see about 90-fold
on off control which we're
pretty excited about. In later experiments
and for further optimization, we've actually gotten
this above 100-fold which is, you know,
about at 200 magnitude level, and so we think we have really
good ability to control both basal as well
as high level expression with this particular molecule.

On the right-hand side,
we can see after we withdraw grazoprevir it actually goes back
to baseline. So again, that allows us
to really get good tight in vivo control
over the product. So finally I want to talk
briefly about smart sensors. I mentioned there's different
classes of these that we care about, you know,
obviously the CAR receptor is one example of a sensor that's detecting
a cell surface antigen. We also are interested
in other classes of sensors, you know, for example,
in the gene therapy context we want to try to build sensors that tell you
what cell type you're in. Why is this important? You know, for companies like,
you know, that are developing AB gene therapies,
currently the best way that they have
to get specificity is relying on vector
tropism itself, however, there is only so many ABs
that have been out there. Even for some of the next gen ABs that are trying
to get more specific, it's oftentimes still very hard
to pinpoint one cell type versus another.

So the way in which we can
add on to the specificity is by building promoters that control gene expression
and have those promoters customized
to certain transcription factors that are only active
in your on-target cells and not in your
off-target cells. The way we do that is by looking
at, you know, large libraries of these transcription factors
on the promoters, so what we do is initially using
sort of machine learning or computational techniques,
looking at transcription factors that are highly upregulated
or highly acted in your on-target cells,
and then number two, based off of the transcription
factor binding sites, we can then assemble libraries
of different artificial promoters
with different spacings, you know, different binding set
numbers, et cetera, et cetera, and then we can do
a high-frequency screen, basically where we barcode
those promoters, and we can then characterize
for activity in the on-target
versus off-target type cells.

And we've done this
in a variety of settings. The collaboration
with Spark Roche is done in a nononcology
setting, but in our internal pipeline. We've been interested,
for example, to build gene therapies
that can target cancer cells versus healthy cells, and so here is an example
of this where we, you know, ran that screen
after the design process, you know, basically tested
in parallel about 10,000 different
synthetic promoters at once. I'm just pulling two examples
out for you here just to showcase. On the left-hand side is a
cancer cell synthetic promoter, that it has about 1,000 times
more selectivity for breast cancer cells
versus healthy cells. We're quite excited about this
because, you know, existing, you know, quote, unquote,
selective promoters that you see in literature,
maybe at most five to tenfold, so we're talking about,
you know, a couple orders of magnitude improvement. We can also design promoters
that have the opposite polarity, which is shown
on the right-hand side, which is about 1,000 times
more selective from the healthy cell
versus the cancer cells.

How can we use
this sort of technology? Now that we can express payloads very selectively
in cancer cells, for example, we could use
the left-hand side promoter, the cancer cell
synthetic promoter, to drive, like, really
potent pillar genes or the expression of cytokines
or other immuno payloads in the true mircoenvironment
in a very selective way and not have those
payloads expressed in a healthy cell context. So this is one area that
we're quite excited about, and then in the context
of other gene therapies, we can use this sort
of promoter control to only express gene
therapeutic payload in the cell types
you care about. So with that, why don't I dive
into a little bit about how we're thinking
about translation? So great, we have this part set
of libraries, you know, really we believe
quite exciting for a wide range
of applications, but now we got to really do
the hard work of turning this
into a real product.

So we spent a lot of time working internally to
figure out, "Okay, where do we apply
these logic technologies or the multiarming
or the regulator dial, the smart-sensor technology? We really want to try to solve
key clinical challenges." So we spent a lot of time reaching out to, you know,
commercial partners and really understanding
what are the major pain points that they're dealing with. Number two,
talking to clinicians to really try to understand, "What are the limitations
with the existing drugs?" And from that, we really
centered around sort of two
major thematic areas. One is oncology. So in oncology, we think
there's a major challenge in being able
to target cancer cells while avoiding healthy tissues,
certainly with chemotherapies, even with conventional
CAR T CAR NK products, ADCs, antibodies, et cetera. The vast majority of cancer
therapeutics are still targeted against
a single antigen, and so now that we can go
after multiple antigen, we believe that could open up
a much wider range of innovations that can go after
that may have been too hard to thread the
therapeutic window needle on.

So we're applying that logic
gate technology I described earlier to a variety
of these indications starting with PML, which Gary
will tell you about later, which we think is one
of the ideal places to start, but that same logic
gate technology can be expanded
to a solid tumor. So for example, you have
a program going after colorectal cancer, again, trying to kill
the colorectal cancer cells while sparing the healthy colon
epithelial cells as well as other
solid-liquid tumors. Second category, we know
it's solid tumors again. There's an immunosuppressant
microenvironment in those
functional contexts, so we not only want to make sure
our CAR T cells are as strong and as potent as possible,
but we also want to be able to recruit other aspects
of the immune system, and that's where that
multiarmy concept comes in. We want to be able to express
cytokines and chemokines that can really call other
immune cells into the fight, you know, endogenous T cells,
dendritic cells, B cells, et cetera,
into that environment.

There we're starting off
with liver cancer. We think it's, again, an area
where we know the CAR T cells can get there. We know we can get
really good control over therapeutic payloads
in that environment, and we know some
of the key mechanisms that are immunosuppressant
in that nature, but again, there are many
other solid tumors that can be addressed with it. So those are really, in essence,
what we're focused on. Along these lines, we are
building internal manufacturing as well as process development
so we can make sure we can actually not just do
the early stage R and D, but we can actually really,
physically make enough product to satisfy patients, so that's really important from
a company-build perspective.

On the bottom half
of this slide, we have worked on partnerships
with other pharma companies, and so this is in areas
outside of oncology, so we really want to try
to deliver our gene circuits, have them reach the broadest
number of patients, and so we've decided
to partner up with companies that have really good
disease-area expertise, who have manufacturing
infrastructure already so we're not waiting
to just do the research and then hopefully
someone can make it, you know, Spark and Bluerock
have invested quite a lot in being able to make AV
and IPSC-derived cell therapies.

So with Spark, our collaboration
is really centered around promoter design
for gene therapy applications in the eye and brain
and in the liver, and for Bluerock,
it's really around building in greater control
in IPSC-based cell therapies and also to be able to regulate
those sort of products. So with that, I'm going to turn
it over to Gary now to tell you a bit
about our internal programs and then some considerations
we have in terms of trying to push
these technologies forward. -Thanks, Tim.
Hi, everyone. My name is Gary Lee. I'm the CSOSMP, and first off, I want to thank
Syn-Bio Consortium for giving us this opportunity
to share some of our work here, particular to Dave and Julie
and Michelle for all their help with organizing
this wonderful meeting. As Tim mentioned, I want to take
the last 10 minutes or so to tell you a little bit more
about the specifics of our logic-gated program
that we've designed for AML, a program that
we call SENTI-202. But before I drive
into the program itself, I do want to start
the discussion and address, as you heard from Tim, all of our pipeline CAR program
will be based on NK cells, so I do want to take
a couple slides and talk about why we believe
NK cells are the ideal modality for the gene circuits
that we're developing.

So on this slide, we're sharing
some of the findings that the field had understood and recently
discovered with NK cells and why they're a potentially
good alternative for T cells. So first and foremost, if I
focus your eyes on the right, there have been really
recent successful demonstration of clinical efficacies
and safety of CAR NK cells which really provide validations
that engineered NK cells can potentially kill
cancer cells in patients. These clinical examples include,
really, the first in human,
SENTI-19 CAR NK trial from MD Anderson
by Katy Rezvani's group as well as in the past year, subsequent confirmatory
finding from Fate Therapeutics induced prior protein
stem cells derived in K
cell pipeline programs across both AML as well
as the LPCL in liquid tumors, truly demonstrating that these
engineered allogeneic NK cells can achieve equivalent
level of efficacies compared to their autologous
T cell counterparts.

So in addition to the fact
that these CAR NK cells can be efficacious, there are actually
several advantages that NK cells
have over T cells. First, because NK cells
did not cause graft-resistant cell disease,
of course because they don't express
anti-HLA T cell receptors, like T cells, CAR NK cells
can be readily deployed as an off-the-shelf product
and provide broader access to patients without the need
to do gene anatomy to knock out several features
on alpha beta T cells. Also, NK cells naturally target
and kill tumor cells based on receptors, like in sampling under
surfaces of these cells to understand whether certain
type of inhibitory or activated lines over or underexpressed, so they inherently
have additional mode of CAR-independent tumor-killing
capacity that T cells lack. But perhaps most interestingly
from these clinical evaluations of CAR NK cells
has been, I think, observed
that CAR NK cells do not cause cytokine release
syndrome and neurotoxicity that are commonly found
with CAR T cells because they just inherently
do not secrete the type of protein
inflammatory cytokines, such as Io1 and Io6, that has been identified
as the main culprit for CRSes as well
as neurotoxicity.

So for these reasons, many believe that CAR
NK cell products potentially can have
a significant eventual advantage in adoption for patient use
as long as, of course, they can achieve the equivalent clinical efficacies
relative to T cells. Next slide, please. So to achieve that goal
at Senti, we developed our CAR
NK platform to address one of the main
concern regarding NK cells which is their potential
to persist in patients, and our solution
to this challenge is our proprietary calibrated
release IL-15 constructs. So with NK cells, there is now
overwhelming both preclinical
and clinical data that suggest that overexpression
of Interleukin 15, or IL-15,
is critical to prolong CAR NK cell persistence
and tumor-killing functions.

So very interestingly, though,
IL-15 actually naturally signals in both in a sys-acting or autocrine-like pathway
as well as in trans or paracrine-like pathway. So at Senti, we leveraged gene
circuit and synthetic biology expertise
and created what we call a novel calibrated release platform,
and the way this work is that we take advantage
of the native proteases that are already expressed
on the cell surface of NK cells. We can design and engineer
a ligand that actually tether
the IL-15 to the membrane and engineered
this chimeric protein and control exactly
the amount of cytokines that are membrane associated and the amount
that are fully secreted, so this concept is illustrated on our left-hand side
of the slide. So what this platform
allows us to do is truly optimize
the IL-15 signaling and get the best
of both worlds which is to have
sufficient membrane-bound associated IL-15
to promote persistence of our engineered CAR NK cells
in a sys-acting fashion, similar to the membrane-bound
IL-15 system that other have described. But in our system, we also
can stimulate and produce fully
secreted IL-15 to more naturally stimulate
our engineered cells in trance as well as other immune cells that are in the patients'
tumor microenvironment in a paracrine-like fashion.

So internally we directly
compare our calibrated release IL-15 protein against wild-type fully
secreted IL-15 to assess whether NK persistence can be
improved with this molecule by performing what we call
a serial killing assays, and that data is described
on the right. So in these types of study,
you can mix the tumor cells and the engineered NK cells and look for inhibition
of tumor growth as well as the reduction of
tumor cells on the cell surface using your intersite
instruments. And so what you see here is that
when we have the control which is tumor cell
shown in yellow, the tumor cells continue to grow
in the absence of NK cells. But as you can see, while both
types of IL-15 CAR NK cells, shown in pink and teal here, can effectively kill
C tumor cells in that very first round on
the left-hand side of the graph, upon repeated challenges when we
actually add more cancer cells back into the system
in the second round and the third round
of challenge, only the calibrated release
IL-15 CAR NK cells shown in teal
maintain persistence and continue to effectively
remove the tumor. So we're really excited about
these results and this novel platform and this technology
we've incorporated in all of our CAR NK programs.

Before I move on, though,
I think I would be remiss not to mention that one of the great
scientists we have, Tom Warren, who is the codiscoverer
of IL-15 and longtime scientist at the NCI recently passed away,
so none of this work, of course, would be possibly
without the giants of the past who paved the way for us
to build new technology for patients who need them. Next slide, please. So, okay, let's dive
into the SENTI-202 program. So this is a program that
I'm truly excited about because I do believe
it fundamentally will solve a number of challenges
for cell therapy in AML.

In this space of AML, despite the recent advances
in AML treatments, the current 5-year survival rate
remains below 30 percent, and allogeneic bone marrow
transplant for decades now, unfortunately, remains the only
potential curative options for many patient
who have suffered from relapse in the relapse
or refractory settings. So the reason for this
poor outcome is really shown
on the left here. First, unfortunately for many
patients, even for those who initially respond
to frontline chemotherapies, most of them will eventually
relapse in 12 to 18 months, and these relapses now has been
learned to be primarily driven by a subset of AML cells
called the leukemic stem cells that are often not targeted
by most therapies today. And the second challenge for AML
is that these AML tumor cells lack a true
clean tumor-specific targets, so therapies that are designed
to eliminate the AML blast as well
as the leukemic stem cells often will put patients at risk for the on-target
off-tumor toxicity, particularly any proposed
specific target for leukemic stem cells are also found in bone
marrow hematopoietic stem cells.

So the Senti gene circuit
solution for these challenges are to develop what we call
an OR logic gate gene circuit to broadly target all subsets
of AML tumor cells, including the critical leukemic
stem cells, to promote a durable remissions, and hopefully the patient
can achieve without relapse. And second, we will use
the NOT logic gate gene circuit the team have already
described earlier to spare the critical
hematopoietic stem cell compartments
in the bone marrow and mitigate potential
wide-threatening toxicity when targeting
these leukemic stem cells. So with this feature, the aspiration
for the SENTI-202 program is to potentially provide
a patient a curative treatment in the absence
of a bone marrow transplant. Next slide. So before I dive into
the specific molecules that we have designed,
I just wanted to show you a product schematic
for the SENTI-202 program. So SENTI-202 again
is an allogenetic off-the-shelf CAR NK product designed and engineered
to broadly target both the mature AML blast
and the AML leukemic stem cells using an OR gate CAR approach
by tightening two different AML antigens simultaneously,
and they're FLT3 and CD-33.

The other component is that
we will intend to prevent on-target off-tumor toxicity
against the healthy bone marrow using the NOT logic gate
and targeting a protein that is specifically expressed
only in the hematopoietic stem cell compartment, and I will describe
that a little bit more later. And finally, as I mentioned
earlier, to maintain CAR NK persistence
and sustain the tumor joint, these cells will also
be engineered to secrete a calibrated released
protein as well. Next slide. So Tim already described
really now the NOT gate work, but here is a slight revision of the slide
that Tim already shared, but in the context specifically
for the CAR NK program we have designed for AMLs. So first, about OR gates,
we have ran through the design, build, test line that Tim described earlier
and developed a bivalent CAR that's designed to
broadly target AML cancer cells.

So this bivalent CAR basically
is similar to other conventional activating CAR
that you may be familiar with but actually contains both the single-chain
barrier binder fragments to both FLT3
and CD33 simultaneously, and as you can see on the left,
this engineered CAR NK is designed to be able
to recognize tumor cells that either express the FLT3
antigens or the CD33 antigens, of course,
when they express both as well.

So as shown on the right, similar to the figure
that Tim already showed earlier, the NOT logic gate as it applies
to the SENTI-202 program is intended to protect
the healthy hematopoietic stem cells, and what we have done is that
through a bioinformatic approach followed by subsequent analysis
of primary samples for both healthy CD34
as well as leukemia cells, we have identified that there's
an antigen called endomucin that is expressed
on the primitive long-term repopulating HSCs
but absent in AML cells. So this can be then, of course,
leveraged to build the inhibitory CAR
shown in purple here. Tim already described
the mechanism of how this works, so I won't dive into it further, but suffice to say the goal is
to achieve a CAR NK product that can broadly recognize
the AML cancer cells while when they see
this endomucin expressed in healthy stem cells, the inhibitory CAR we get suppress the effect and function
of the NK cells and protect those cells.

So one last important point
I do want to make on this slide is that of course because we are
protecting healthy HSCs, we do not need to protect
100 percent in every one of them
because based — There have been decades
of transformative experience, including autologous
bone marrow transplant where physicians that had known
for a long time now that as little as one 1 percent
of the patients' own hematopoietic stem cells
is sufficient to be populating and generate
the entire immune system. So in conversation
with many AML KOLs who were transplanted as well, the challenge they have provided
for us is to then protect at least 10
to 20 percent of HSCs in every patient
which they believe would be clinically meaningful. Next slide. So in the interest of time, I won't share all the data
we have in this program, but we have certainly optimized
the SENTI OR gate CAR
construct against AML. We have plenty of in vitro data
shown killing of AML tumor cells as well
as primary AML patient cells, but that data was
presented earlier at the AHCA meeting in May.

But here on this slide,
just showing that, of course, our OR gate or CAR NK cell can show improvement
in the tumor response and xenograph mouse model. We've done a few of them. Here is showing the MD411 model. You can see that the OR gate
or CAR NKs on the bottom row on the left
significant suppresses and reduced tumor burdens, and on the right showing
that associated with that, we have improved survival
advantage in these mice as well. So last data slide
on the next slide, just showing you about the
NOT logic gate data.

Here we have constructed
endomucin-specific iCAR to selectively shut down NK
killing functions against cells that express
this particular safety antigen. So what you're seeing here
on this slide is that we've engineered NK cells
to express the aCAR, but also if an iCAR,
and then coculture them with either
a FLT3-positive leukemia cells and also a model cell where
this cell line also express the antigen endomucin.

So for the control iCAR on the
left which targets the HR2 antigen
is a mistarget of iCAR. You can see that these CAR
NK cells indiscriminately kill
both cell types, the gray
and the teal as expected, and when we actually
put in a functioning endomucin iCAR from the two groups
on the right with two different
I10 domains inside, you can see that
we're able to confer 50 to 70 percent of protection
against cells that expressed
an endomucin antigens. So this, of course, you know, we're really excited
about this data exceeding the 10
to 20 percent of the target that we initially set, and to the best
of our knowledges, the first demonstration
of that function did not occur in CAR NK cells. The company is now full speed
ahead in trying to move
this program into the clinic.

Next slide. So I think when Tim and I
decided to tag team this talk, I think Dave and the organizer
asked potentially whether we can have
a discussion about, you know, different
translational aspect on how to take potentially novel
synthetic-biology technologies and transfer them into patients. And so a little bit more
about my background: I've had now 20 plus years in
the cell and gene therapy space. Prior to joining Senti 3 years
ago, I spent 13 years at Sangamo Therapeutics,
where I was part of a team that helped translated
five different genome edit cell-therapy programs
into the clinic across different disease areas, including oncology,
infectious disease and hemoglobinopathies and modifying different immune
cells, including T cells, NK cells as well
as hematopoietic stem cells. And so, you know,
I think one component about, you know, my experience in how
to leverage novel technologies and take them into the clinic is that it's a clinical program
for evaluations.

Obviously, there's — Each program is different
and unique in its own right, but I'd like to
at least summarize into four different categories and, you know, maybe raise some
points for your consideration and maybe discussions afterward. I think first and foremost
within in INB that there's
three major components. There's the research and
the preclinical aspect and the pharmaco
and toxicology sections, and so for these
particular areas, I think one aspect
is that really, I think for many,
many of the attendees here, we are in the space of discovering
very novel technologies, and there's no one
who better understand the potential risk benefit
of this technology that we provide the patients
than the sponsor themselves. So it is really critical to
communicate their aspects in terms of the potential risks,
in terms of talks, et cetera to the agency and explain how you may plan
to assess these risks. And so one example
I can give is that, you know,
almost 15 years ago now when we first started
with the genome editor T cell program at Sangamo
as the first in-human program, really the concept of making
a double-stranded break in the human cells
for therapeutic application is quite unique and novel, and so the team went through
a very thorough analysis and proposal to the agency on how we intend
to assess those risks.

Over the years, the type of
methodologies that improve and that tie,
for instance, back to that, but I think it's really critical
for the sponsors who discover these technologies to clearly communicate
and understand the risks of toxo termogenicity
to your technology. I think an important part
is understanding the limitation of preclinical models,
to understand, for example, in a xenograph model
that does an immune system, how and what limitation it has
on the immune cells in terms of and graphing, establishing
and in other aspects of that. I think the second point
I do want to discuss is CMC because for many
of these advanced therapeutics, it often involved
multiple products. For cell and gene therapy,
it involved multiple steps and oftentime ancillary products that is used to generate
a final cell product or gene therapy program, and these would,
of course, potentially include even upstream program products, like for body's productions
and infectors, et cetera.

And so understanding that
how each of these components can impact your final product
is critical, and because of the nature
of these multiple products involve along the way,
from a time line perspective, in my experience this is often
the rate-limiting steps to a clinical evaluation
of your program, so understanding
the time line impact on CMC is critical for the development
of your program. And finally, I don't want to
speak too much on clinical. That's definitely
out of my area, but just from my experience,
that it's never really too early to think about development
of the clinical protocol, starting early,
having discussion with clinicians in KOL who treat patients
on a daily basis to understand
the patients' need and how this technology
can improve patients' life, and it's really critical.

And the idea of having — really centered around having
the patient centric in mind where you can design
the studies that are friendly to the patients and work
with clinicians to make sure that you can explain
these advanced technologies to patients in a way
that everyone can understand. And finally, on the bottom
right, certainly we cannot
evaluate this, the programs
and the technologies, without the help
from the regulatory agencies. There are many different
regulatory meetings that one can arrange
with the agency, including the interact meeting
as well as a pre-MD meeting. For Advanced Therapeutics,
I was fortunate enough to work with many, many experienced
individuals at the office over the years,
and really clear communications and using these pre-IMD meetings as tools to communicate
with the FDA so you can have clear
plans and understand on both sides
that IMD is critical.

I think last point
I want to make: One thing I learned is that it's
really helpful for any sponsors to keep up communication while with the agency
to stay organized and keep track of all your
communications with the FDA. So with that, I think we did
actually use up more time than I intended, Tim,
so maybe we can still have time to hopefully take
a few questions, though. -Yeah.
I think maybe, Gary, just reading the questions here,
thank you for them. I guess one question just in
terms of general advice on reaching out to the FDA on early phase
clinical-trial designs? -Yeah. I think, you know, I believe
there's probably an FDA representative
at this conference as well, so just from my experience,
first of all, the agency have clear guidances
on these types of meeting that they provide
to scientists and sponsors who want to run
clinical trials, so definitely use those
resource and understand what the agency
can help you with. And from my experience,
there are now two meetings that you can potentially
arrange with the agency prior to the IMD filing.

One of them is
the interact meeting, previously called
the pre-pre-IMD meeting some time ago, but really this is a tool
that one can use to have an early discussion
with the agency if your technology is novel and so you can get some guidance
on how you should think about the pharmacology
and toxicology aspect as well as, you know,
different components as well. And finally, the FDA will grant
a pre-IMD meeting to discuss
what your IMD plan is, and I think it is important
to think about the time line to arrange this meeting. It is a balance between
how clear a path we have already developed so you
know exactly what you're making so you can provide
a briefing package that is clear
and understandable to the agency and can add very specific
questions and then get feedback so that you can execute
those studies in agreement with the agency
in the IMD submission. -I was going to
just chime in quickly.

I know we're at time,
but we're going to eat a little bit into the break because I think you have
some great questions and great potential
discussions, so maybe we can have
5 more minutes, but we specifically have some
FDA representatives here, so I wonder if
they want to chime in and maybe compliment
what you just described, Gary. And if I put them too much
on the spot, we can also hear from them
again in the afternoon. [ Dog barking ]
-Hang on. That's a problem
I wasn't prepared for. No meeting would be complete
without a barking dog. This is Karen Elkins from CBER,
and Carolyn Yong, my colleague, is probably
a little better prepared to respond
to questions, but we are going
to cover interactions and some of the specifics this
afternoon in the presentation, so hold that thought.
-Perfect. That's a perfect teaser,
and just want to get — It seems like you
can see the chat.

If you need my help to moderate,
I can, too, but there was one that I was
really hoping you can get to because I know we have
a lot of early career and trainees in our audience. If you could talk about, Tim,
maybe your perspective from that
academic-to-industry transition and maybe what they should
be thinking about if they're considering startups
or industry? -Yeah, for sure. I think it's been an interesting
experience for sure.

I think a couple of things that I've learned
throughout the process: One is, you know,
at least personally, I came very much from
a technology-centered view of synthetic biology. And, you know, I think
one of the things that was quite important
early on is to spend a lot of time
up front making sure that you're
applying it to a problem that people care about
from a clinical perspective and not just necessarily
be enamored with the cool technology
you've developed, just really spending time
with the clinicians, drug developers,
others to really make sure that what you're, you know,
able to try to build has a really sort of important
pull from the environment because that would just
make life a lot simpler from a downstream
clinical perspective, fundraising, et cetera. Two is, you know,
I think the start-up world, you know, the biopharma world
is very much a team sport. It's very rare for
any one person to have all the skills needed because it's just such
a multifactorial problem, right? You need people who know
regulatory, who know manufacturing, who know
the signs, who know clinical, and then there's obviously
the business side of things.

And so I think building
that team early on of people who have quite
complementary expertise and can work together with you
is super important, so — And it's not easy
to find people like that, right? I think you got to spend
a lot of time networking, you know, working through
your own, you know, so people that you know
and reaching out and just being, you know, forthright
about doing that.

It's a lot of sort of blood, sweat goes into
that whole process. So, yeah, that's been
my general experience there. It's been quite exciting,
though, to see technologies be able
to pull out of the lab and, you know, start applying
to specific areas going forward. Gary, I don't know
if you have anything to add to that from your perspective
in terms of your own career. -Yeah, I think I was just
reading the chat, as well. I think certainly, you know,
we love — I think some of those questions
discussed from the agency are a much better
guidance to them, so providing directions and how to reach
our interactive agency.

But maybe I can speak
on my be — on the side of the sponsors who want to evaluate
their technologies in the clinic and just my experience
on what can lead to a productive conversation
with the agency. I think when you have
interaction with agents from my perspective, it's actually much more critical
for you to understand that the agencies
are not your advisor, right? They're not there to tell you
what to do. It is important for you
to basically understand what exactly you're trying to do
and give a proposal to agency and provide them
with all the information they need to understand
your proposal, have all
the scientific information so they can make
a judgement call and provide support
to let you know what you're suggesting
is reasonable or not, right? And so I think
open-ended questions tend to not to be
very productive. It is important for the sponsor to be very clear
about what their goals are and provide all
the scientific justifications so that agency
can make a decision on whether your proposal
is accurate or not. -Those are all the questions, the majority of the questions
I've seen.

I guess there's one more
on monogenic disease. I think I briefly mentioned
one of the challenges with some monogenic disorders that they may only occur
in certain tissues, for example, like certain cells in the brain
or certain cells systemically. And so really, they were
very much focused on building promotors that target
those specific cell types to give you that level
of selectivity that you want. So hopefully I've addressed
most of the questions I'm seeing here, and I'll turn
it back to Michelle. -That covers what I saw, too. It seems to have a lot of
really positive feedback and enthusiasm
about your presentations. So thank you both
for being here today. This has definitely
been inspirational. We hope that you're here
for a little bit longer because I can imagine you'll
have some more chat questions that keep coming up, but really grateful for you
sharing your experience. -Thank you.
Welcome back from the break. For this session, each speaker
will give their talk, and then there will be
a panel-type Q and A at the end.

If you would like to post
any questions during the talk, I will keep track
of the questions, and they will be addressed
and who they were addressed to. Each of the following speakers
does research and engineering mammalian cells or the cell itself
as a therapeutic. The benefit of cell-based
therapeutics in adaptive response, where the engineered cells
can target and adapt to changing conditions
within the body, the common challenges
to each of these approaches is reliable detection
of the disease state and transduction
of this detection into a therapeutically
beneficial response.

While the mechanisms for
detection and response differ, control and specificity must be
addressed with any approach, as you will see,
with each approach having its own advantages
and considerations. With that, I'll hand it over
to you, Parijat. -Thank you for the kind
introduction, Nicole. So I'm going to share my screen. Okay, so again, thanks for
the kind introduction and also thank you
for the organizers for giving me a chance
to present our work. So I would like to start
with giving you a scenario for imagining a medicine that
assembles where it is needed, when it is needed
and without passing through the heavy tissues.

And that is the concept that we are trying
to bring to the clinic. Cell-based diseases evolve
and transform and have learned to evade
the immune system, like, similar to, like, cancers, viral diseases,
autoimmune disorders. These are like living diseases. In order to respond to these
living diseases, we have come up
with a living medicine, the T-cell biofactory,
and this is how it works. The disease cells, they express
biomarkers on their surface which can be identified
by T cells if they have
a chimeric antigen receptor. And this is something
which is T-cell biofactory is, but when the chimeric
antigen receptor engages the disease cell, it turns on an intracellular
machinery in these cells which we have hacked into
to express a protein that will neutralize
the pathology that has triggered the T-cell biofactory
in its first place. The first use case scenario is
an antitumor T-cell biofactory. The extracellular matrix
is a barrier that prevents entry
for antitumor agents into the tumor microenvironment, including the T cells
themselves, and that turns the — that's what
is responsible for — one of the things
that would be responsible for turning tumors
into cold tumors.

The T-cell biofactory
is a T cell that has been engineered
to express enzymes that degrade
the extracellular matrix. The first graph here shows
the production of these enzymes. This is a base level expression
from the control CAR T cells, and the other two
are different versions of the T-cell biofactory
that we have developed, and this version six is looking
so far the best one. And this graph here shows
the activity of the enzymes that were produced
from these T-cell biofactories and show that these enzymes
are expressed only when the T-cell biofactory
is active and, again, of course, are not there from
the control CAR T cells. And these enzymes
can be inactivated, as shown in this,
if we use enzyme inhibitors. This work was in vitro — or sorry, yeah, in vitro,
and we did some in — ex vivo work, where we extracted
the tumors from mice and treated them with
the enzymes that were produced either from the control CAR
T cells or T-cell biofactory, and we see
a big significant difference between the
ECM degradation product that was released from the — in the two
different experiments.

So far the data that I showed
you was in Jurkat cells. However, it is important
to transform this into primary T cells
in order to make it
clinically relevant, and so we put
these genetic circuits into the primary T cells. And this graph here shows
the enzyme activated that is produced by
a different T-cell biofactory compared to the control
CAR T cells. And the IO2 and IFN secretion, which are the markers
for the T cell activation, were compared between
the control CAR T cells and T-cell biofactory
which were very similar, as you can see from the graph.

The cytolytic activity
of the T-cell biofactory was also very similar to that
that was observed with control CAR
T cells almost overlapping and, of course,
very significantly different from what was observed
with the unmodified T cells that do not have the specificity
towards the antigen. And the data there so far
has shown us that the T-cell
biofactory performs as good
as control CAR T cells except that it has
the capability to destroy
the extracellular matrix. And hopefully, as we transition
into better animal models, we should be able
to turn cold tumors hot. The second use case is
an antiviral T-cell biofactory. So it has been found
that interferons, the — specifically type
one interferon, is abrogated
in the COVID-19 patients that progress
to severe disease stages. So we generated a T-cell
biofactory that expressed interferons, type three and type
one interferons, both of which are antiviral. And these interferons
are expressed only when the T-cell biofactory
engages the infectious — the infected host cells, the host cells that have been
infected with SARS-2 virus.

The production of the
interferons from this biofactory has been found to protect
the host cells during — by turning on the innate
antiviral mechanisms at the prodromal stages. The top row here shows
the interferon activity when administered
before the viral challenge, hence a prophylactic activity
of type three and type one interferon. The bottom row here shows
the therapeutic activity of these interferons which means
they were administered after the viral challenge. The dotted black line here is
control from pure interferon beta that was obtained
from a commercial source, and we can see here
that other — the interferon that is from
other biofactory that is — might be the same or better compared to what was
commercial interferon.

Again, the difference really is
that the commercial interferon has to be
administered systemically because it's soluble interferon. However, the interferon
from the T-cell biofactory is produced only when the
biofactory engages the target, the infected host cells. The next few scales is where
we transform these T-cell biofactory, which we have named it
differently, into cell-based diagnostics, and there are
two clear advantages. We develop antigen tests
and serology tests with this, and there are two
clear advantages for this. One is that these are based
on cells which are self-replicating, and so they are easy
to manufacture. The second advantage
is that the job of these cells in our bodies is really to find the diseases
among all different mix that is there in
physiological environment.

So you don't really need
to purify any — do any sample preparation
before we use these cells, so hence they are
very easy to use. We developed antigen tests
and tested them with oral swabs obtained from infected mice
without any sample purification. The schematic here shows how
the antigen tests work which is similar
to what I described before, that the infected host cell
is presenting a spike protein on a surface,
and the T-cell biofactory — Here, we call it
diagnostic cell — has a chimeric antigen receptor which, when it engages
the spike protein, expresses that protein.

These two graphs here show
that the mouse oral swabs that were obtained
from infected mice, and these mice were tested
with GPCR to be positive, were all also detected
with our diagnostic cell, and the negative control
did not show any background. This graph here shows
that we were able to reduce the reduce the time of our assay
between 1 to 2 hours. However, there is significant
room for improvement, and with more work, we should
be able to drive it down to — Theoretically, we should be able
to drive it down to the time it takes
for a single cell to inform.

But, of course, this is — It could be somewhere
in that ballpark. -Two-minute warning.
-Okay. Another unique property
that we observe from these cells is that these cells can directly
detect viral particles which is unlike what these cells
do because the T cells — these T cells are expressing
chimeric antigen receptors which cross-link and initiate
a similar activation cascade which is otherwise initiated
by a chimeric antigen receptor when it engages
the antigen-presenting cell. So this is something different
in these T cells, that they are able to detect
the viral particles directly.

The serology test is dependent on two different
genetically engineered cells, a pair of genetically engineered
cells rather than one cell, and reporters here which is,
again, a diagnostic cell — The whole thing we call
diagnostic cell complex. So the reporter cell here is
expressing a reporter protein only when the immune synapse
is facilitated with an antibody which it is able — which — towards which
the test is very specific. So what is important
to note here is that we are able to detect
soluble antigens in the environment,
and as we progress in our work, this is one thing
we might be able to transform, a human response
into a therapeutic response by using such T cells into —
when used as therapeutics.

The graphs here show that
from human serum, all COVID patients
that were detected positive with QPCR in the past
were also detected by our diagnostic cell
complex to be positive. And control subjects
did not show any signal, despite the receiver
operating curve shows that, at the sensitivity of 97 person,
we had a specificity of 93 — -Time has expired.
-Okay. Looks like I'm out of time,
so in the interest of time, I'm going to skip on our work
on in-line micro-electroporator that rapidly generates about
2 billion cells per minute, and most importantly,
our acknowledgment team that has diligently worked
with me to bring this technology
to fruition, and, of course,
to funding agencies.

And we are excited about
our work, and any of you guys who are looking for a
postdoctoral scientist position, please feel free
to contact me. Thank you. Thanks. -Okay.
I'll go ahead and start here. So first of all, thanks
very much to the organizers for putting together
this fascinating meeting, and it's really
exciting to see how these exciting
sophisticated cell therapies are moving
from bench to bedside in, for example,
the work Tim and Gary showed at the last part
of this session here. So today I'll be talking about
some work we're doing to develop some clinically motivated
platform technologies to enable a next generation
of these advanced cell products. And like many people
at this meeting, we've been motivated by
the vision of trying to realize design-driven engineering
of cell therapies. And in addition to making
these products possible, a real advantage
I want to highlight is that getting good at this
can enable us to spend much more time
and money evaluating new ideas rather than trying to build
and tweak prototypes.

And so the technology I'll be
telling you about today is really motivated
by that perspective. We've been inspired
by the benefits of CAR-T cell therapy, and the talk I'll talk to you
about today, though, is motivated
by also a key limitation of all CAR-T cell
approaches described so far which is they require
the knowledge, the foreknowledge,
of a tumor antigen. And there are many cancers
for which there simply is no unique tumor marker, or we just don't know
what it is in certain patients. So there's a big opportunity
to extend these kind of benefits to other groups
of patients and groups. And so the motivating
application I'll mention today is trying to build
tumor-specific therapies that recognize the tumor not by
engaging with the tumor antigen but by sensing one
or more combination of features that distinguish
the tumor environment from other parts of the cell. And so what we've been working
on for the last couple of years is building the technologies
required to do those steps. So I'll talk to you about
each one of those in turn here.

When you start with a big
therapeutic mechanism like this, you need to break it
down into parts, and so this is a now canonical
synthetic biology paradigm of thinking about those parts
of the sensors, processors and actuators. And when we started this work, the first gap we realized
was a lack of suitable sensors to carry out this kind
of specific engineering. A number of years ago now,
we invented a platform — a receptor platform called the modular extracellular
sensor architecture. This is actually invented
by our session chair, Professor Daringer. And the way this system works
is, there are two protein chains, and upon ligand-induced
dimerization, intercellular trans-cleavage
event releases a transcription factor
and enables you to regulate either androgynous
or trans genes. And just to illustrate
why that's important, we showed in some other work
that when you implant this, for example, in a human T cell, you can build new input-output
functionalities.

For example, we showed you can
engineer a T cell to sense VEGF and, in response,
secrete IL-2. And the reason that matters
is that VEGF is a normally
immunosuppressive cue. IL-2 is a normally
immunostimulatory cue. There is no cell in the body that exhibits that input-output
relationship. You can come up with many
different ideas for what might make
for good input-output pairs, and our goal really is to make
it possible to implement those and then try them out
in animal models. The fun thing about this yield
is that it's now not the case that we're limited by receptors. In fact, there's a plethora
of different systems now which all have
different features and different benefits
that they offer. This is a figure derived
from a forthcoming review in "Nature Chemical Biology" that we cowrote
with Leonardo Morsut, one of the coinventors
of the SynNotch platform. And one of the points
we make here is that although there's a lot of power
in each of those approaches, it still remains challenging
to figure out how to put these pieces together to create
a combinatorial product.

And we actually are starting
to learn a lot about why that is
in the synthetic receptors base. One of the barriers
is knowledge. Oftentimes, when you try
to employ a nature part, it turns out that
the prevailing wisdom about how that receptor works,
for example, is wrong, and you need to build
new understanding to be able to use those, even in the constructs
that are synthetic. Burden is a feature where if you
engineer overexpression of certain things
in a cell, those expression limitations
can start to interact and cause your system
to behave in unexpected ways. And probably the most
important contributor to lack of performance
in sensors is variation
between individual cells.

And really, we mean variation
in the levels of expression of your components
as well as in cell state, for example. So how do we go about
tackling that? Well, one idea we came up
a couple of years ago was to build a receptor
that is inherently less subject to this idea of susceptibility
to variational failures. And so this is a mechanism. It's very similar to the cartoon
I showed you before, except in this case,
the signaling occurs via reconstitution
of a split protein. When we first tried this, it was clear that that protein
was just coming together way too fast all the time,
and we realized we had a problem where we needed to tune down
that level of reconstitution to make it work
with our proposed mechanism.

And upon investigating,
we realized this is actually
a general challenge in the area of using split
proteins for bioengineering. And there is no general solution
to that idea of tuning down. You can do screens, but that
doesn't really work for a lot of applications. So we teamed up
with Srivatsan Raman, a computational protein expert,
to build the new methodology that enables us
to solve this problem. And I'm just summarizing at a
high level what this algorithm, this process called SPORT,
enables you to do. It employs computational protein
to design to really hone in on a very small set
of protein variants that you can build
and screen individually.

Something like a dozen
is probably enough. And they scan through that space where reconstitution is
very easy down through very hard to find the Goldilocks zone. And when applied to this
receptor engineering, we were able to take an
initially nonfunctional idea — The black and grey bars
at the left here are initial constructs. It does nothing in response
to ligand. And it enabled us to build some of the highest-performing
receptors characterized to date. That would include both
synthetic receptors and natural receptors. And so one of the key features besides performance
that we characterized but I'm not showing here
is that this happens in part because these receptors are not
nearly as sensitive to variations in expression
that you see across cells.

So all the cells
in the population basically are working
as predicted. So the last part of the story I'll extend to the next layer
of engineering which is looking at the need
for processors or building gene circuits
to process information. And again thinking back to the
motivation example, the tumor microenvironment
is distinct but not entirely distinct
from other tissues in the body, and so we expect that we would
need to do something that involves combinatorial
sensing of features to really get specificity and really to get
the kind of specificity we need for a translational
application. When we started looking into how
to do this in mammalian cells, we realized that, at the time, there was really no tool kit for
doing this — years back now. We reported this platform. We built and reported
this platform called COMET which is a big library
of engineered promotors and transcription factors that were all
quantitatively characterized, and so right off the bat,
you have a menu where you can pick a strong one
or weak one or a middle one.

But the story
I'll highlight today is that we went the next step and used fairly
rigorous mathematical modeling to both understand
how these design choices were changing
the performance of our parts and enable us to
start looking forward from explaining the biology we already built to trying to
use models to do forward design. And so in a story we published
a couple of months ago, we tested and showed
that that is possible, and these cartoons
are just summarizing the idea that with that kind
of capability in hand and a few other mechanisms
that we built and modeled to give our tool kit a few more features,
we were able to go to, like, the chalkboard,
really in the computer, and simulate
a lot of potential models, a lot of potential circuits
that, if you looked at it and had some knowledge
of systems biology, you might think
that looks plausible.

And when it turns out that
if you tried to build it using the actual parts we have, the computer predicted that many
of those good-looking ideas were actually not good-looking. But a few of them were looking
really good-looking, and as it turns out,
when we went and then built only the ones
predicted to work well, they all worked well
the first time. And this is really unprecedented
in this space, and we tried to really
break this methodology by building increasingly
complex circuits, more complex than we need,
but really just trying to see the limits of our ability
to do predictive design.

And to a one, it turned out that
if the simulations we were building
suggested that these designs would be useful
or would be functional, they turned out
to be functional. And we think the essence of this comes down to really capturing
how variation impacts the performance
of these circuits, just like thinking about
how variation impacts the performance
of those cells. And, in fact,
these designs that we worked are not
as sensitive to variation, and that means they don't have
to be tweaked and tuned, probably a big part
of why they ended up working. And so I've summarized
the application of this concept to building cellular logic
which is a good test case and kind of builds on a lot
of systems biology foundations that let us test our abilities, but there are
a lot of other things you might want to engineer
yourselves to do besides logic. So we've implemented
analog information processing, amplifiers. We've changed continuous
responses into more discreet responses
using ultrasensitivity and even filtering through
some kind of circuit design. So — -Two-minute warning.
-Perfect.

The cool thing about this is,
we're now at a point where our ability
to do this design far exceeds actually translational goals
we have in a couple of ways. So the last little bit
of the story I'll offer is kind of a preview of where we're going next
with this, back to the original motivation. We'd like to see
if we can integrate these synthetic capabilities with some of the natural
mechanisms that are useful for distinguishing tumor sites
from healthy sites.

And we've been interested
in the last couple years on focusing on hypoxia
in particular as a signature
of many different tumors, a signature that probably
isn't one that tumors can evolve to remove. So if you could
integrate hypoxia sensing with synthetic circuits, this would be a great way
of potentially tweaking the way your circuits
get activated at a tumor site. And one thing we have recognized is that the cell's natural
ability for sensing hypoxia is not necessarily the right
sensitivity or magnitude that you would need to turn
on a cell therapy, and so I'm showing a brief
couple of unpublished studies showing we can build
synthetic circuits, interface them
with natural mechanisms and change the performance
characteristics of those hypoxia-responsive
features to do things that are more useful
for cell-based therapies. And we're in the process now
of evaluating these as well as a couple of other ideas
in some preclinical models, similar to the ones
that you've heard about, and we're looking forward
to sharing those in the near future.

So hopefully I've shown you that
we're at the point now where we really
can do forward design, and especially if we wanted
to spend more time evaluating ideas
rather than building parts, thinking about the downstream
translational challenges in terms of cell
manufacturing, cell variation
and performance in the body. should have a lot of benefits in being able to throw a lot
more good ideas at the wall and pull out
the really excellent ones. So finally, I'll end
by thanking my team. This is a recent photo
of our group, especially Professor Daringer,
who's hosting us today, for inventing the story
that started us on this track, as well as key collaborators Neda Bagheri, Niall Mangan
and Vatsan Raman.

And I'll be looking forward
to your questions when we get there. Thanks. -This is one of my
favorite meetings to go to. And so my lab is very interested in directing hematopoiesis
to make platelets in vitro. And hematopoiesis is a process
where hematopoietic STEM cells, which are here,
live in our bone marrow, and they are
actually responsible for making all of the cells
of the blood system, including all of
our immune cells in addition to our red blood cells
and platelets. So we are really focused on making red blood cells
and platelets because these cells
are anucleate which means they don't
contain a nucleus, and so the red blood cells
actually shed their nucleus as they mature. And platelets actually come
from their progenitor cell, which is the cell
right before it, which is this very large
polyploidy megakaryocyte which has this very
long cytoplasmic extensions, where the platelets actually
blub off of the megakaryocyte. So today I'm going to talk
to you about making platelets in vitro, and you might ask,
"Well, why platelets?" And, well,
that's a great question.

So platelets are actually,
as I said, they're anucleate, and so all of the engineering that we do upstream
of these platelets, none of that exogenous DNA
is present in the platelet, and so that actually gets us
a little bit closer to making this
clinically relevant. We also have about a trillion
platelets in circulation in humans which is, like, really
mind-boggling to me, but also they're very small
but mighty cells. They're about 1 to 3 microns
in diameter, so they can actually go into
a lot of different places. Many of us are very familiar
with the clot formation that the platelets
are involved in, but they also have other
very important roles which includes hemostasis, wound healing, endogenesis
and inflammation. I'll briefly mention some
of these later in my talk. And so for the clinical
relevance for making platelets
in vitro, it includes the fact that
platelets are in circulation for about 9 to 10 days, and so our megakaryocytes
are quite busy, and when our megakaryocytes
become exhausted, these hematopoietic
stem cells actually have to make
more megakaryocytes.

The shelf life of platelets
are only about up to 5 days, and so this means that we need
a constant supply of platelets clinically so that we can
help patients as they need them, and they also need to be
stored at room temperature. This is very significant
from a clinical perspective because this really makes
shipping difficult and in addition
to that 5 days. So what happens when you
actually try to store platelets at 4 degrees, like you
can with red blood cells is, you actually activate them, and when a platelet
becomes activated, so initially, they're filled with hundreds of
different kinds of proteins. They become activated,
so they practically explode, but they release
all of their proteins into this extracellular space, and some of these proteins
recruit other platelets, which then activate
and recruit others, and so this is what we know
in terms of a clot. I don't know why
this isn't advancing. Oh, there we go. So one of the questions
really is, you know, what drives cell fate decisions? There are many factors
that drive cell fate decisions certainly, but one of the main factors
are transcription factor levels, and so we can have — there's a huge body
of evidence in the literature that actually suggests that low
levels of transcription factors, for example,
go in one direction, medium, a second direction and high levels
in another direction.

This is very difficult
to recapitulate when growing cells in vitro, and so we often resort
to media conditions. So what we did to sort of
demonstrate the rationale for using synthetic
Riogene genetic circuits, is we took human iPS cells. It takes about 19 days
to make megakaryocytes. We actually went to STEMCELL Technologies to buy
extremely expensive media and did this
on a very small scale so that we could see,
you know, what is sort of — What can we get with
just media conditions? And we ended up with a
pretty heterogeneous population. We did get megakaryocytes
certainly, and what we did then is, we looked at how many platelets
or platelet-like particles we could harvest
from these cultures. And, in total, we only had
about 8.5 percent of cells that looked like they could be
platelet-like particles or platelets,
and of that 8.5 percent, only about 18 percent,
in fact, were platelets, and so this number
is really low. It's not clinically relevant,
and it's also not scalable because, again,
this media is so expensive. So this is where we believe
that synthetic biology is going to have a huge impact
in the fields of tissue engineering
and cell fate decisions and even understanding
basic development questions, and so what we're doing
in my lab is, we're building genetic circuits
so that we can very tightly regulate various
transcription factors.

They can sense and respond to
the cell type that they're in, but also we can tune,
very tightly tune the level of expressions, and we can actually
push the differentiation to get robust populations
of, in our case, megakaryocytes that
we're looking for. In addition to tuning
the level of expression
of these transcription factors, another important gene
expression element with regard to development is also the timing of
when these turn on, and so it's not
only the level, but it's also kind of when in
that process that they turn on. And so we're building
lots of tools, and so this is
one of the tools that we built when I first started my lab, and it's basically a genetic
circuit, and what we did is, we put this circuit
into mouse embryonic stem cells, and so these cells
grow as these clusters.

There's hundreds of cells
in these individual clusters, and without the inducer,
gene expression is off. We can then turn it on with
the addition of the inducer, but we can also tune
the level of expression, so this is really
important for us in terms of these cell
fate decisions. So then the next question was,
you know, where are we going to —
where do we begin? Where do we start? And so the question is,
first, you know, do we go to the first cell
before this? And these are megakaryocytes. This is a —
in this flow cytometry — and if —
And these are really cells. You can see their cytoplasmic
extensions coming off here. These are —
While they're beautiful, they're not the right cell type
for us to go for because they have — There's only about .01 percent
of total nucleated cells in the bone marrow
are actually megakaryocytes. They're also nondividing cells, so they're not really
a good cell to look for. Our next step is to look at —
kind of move upstream and to look
at hematopoietic stem cells, but these cells, it turns out, actually don't grow
very well in vitro.

This is very consistent
with data in the literature. We found about a peak
around 9 days, and then after 9 days, they start to really rapidly
diminish, and that's — They either
spontaneously differentiated or they would
undergo senescence. So HSCs were not the right cell
for us to go for, so we went all the way
to the top. We went to pluripotent
stem cells, and we thought, "Well, these cells
actually grow fairly easily. They're pretty easy
to transfect." And so we decided to — And they also become
any cell type in the body. So we decided to go there
and then build our circuits to push the differentiation
down into the mesoderm lineage so that we could start
to get our blood cells that we were interested in. We're building lots and lots of
different circuits to do this and to regulate the — not only the level of
transcription factor expression, but also the timing
for it to turn on.

And I have this
little symbol here. I don't have time to talk
about it here, but we have built a device
where we can actually — It's an in vitro device
that we can actually get hundreds of platelets
per megakaryocyte. So we're able to actually use
our circuits to push the differentiation
into megakaryocytes and then get this
huge number of platelets, but the reason why I don't have
time to talk about it is because I want
to talk about a second layer to this programming of cells, which is that we're actually
engineering these cells to be novel
delivery devices as well, and what that means is that — So if you remember
when I talked about — So these platelets come
off of the megakaryocytes, these long cytoplasmic
extensions, so — -Two-minute warning.


So the cytoplasm in the platelet is the same cytoplasm
as in the megakaryocytes. So we actually load up
these megakaryocytes with lots of different
therapeutic proteins, and then we can create our
engineered platelets in vitro. So this is a picture
of a green megakaryocyte. You can — I hope you can see there's,
like, little green cytoplasmic extensions coming off, and then
we put it through our device, and we end up with lots
and lots of different green fully loaded platelets,
and so some of the — This is a platform technology
that we've been developing, and so we have —
We build these circuits to, first,
direct cell fate decisions, and this is a basic science
sort of focus in the lab, so we can go from
a pluripotent stem cell, and we're very interested in looking at
the mechanisms of development, in terms of from pluripotent
stem cells to a hematopoietic stem cell.

There are many patients that
actually have nonfunctioning or misfunctioning
hematopoietic stem cells, so they actually don't develop
the mature blood cells that are required, whether they're platelets
or anything else, and there are a lot of different
genetic issues that take place, and so we can actually
build our circuits to look at not only a known gene
that is causing a problem, but we can also do some systems
sort of biology where we can look at, you know,
RNA seek and things of, what else is going on
mechanistically in these cells that are actually
causing hematopoietic stem cells
to not properly develop? We can take that same idea and
look at the mechanisms of — some patients have perfectly
fine hematopoietic stem cells, but they're unable
to make megakaryocytes, and then also the next
step would be the — into the platelets. Some megakaryocytes don't make
that proplatelets very well, and so patients have
low levels of platelets.

And so a next platform
technology that we are — or a next aspect
of the platform technology — are these really — Are our clinical trajections
that we're currently working on. So one of these —
-Time has expired. -Okay, I have, like —
My time is expired, apparently. I have, like, 30 seconds. So the — Is actually to make
these platelets in vitro, we can do all sorts
of different treatments, where we're also loading up
our platelets with different therapeutics to work on both
metabolic disorders and also targeting —
circulating tumor cells. So with that, I'm out of time. Of course, I have to thank
all of my wonderful students who work very hard
and also my friendly sources, so thank you,
and I'm happy to take questions.

-So the first question
was for Parijat. How does ECM degradation
affect cancer progression, e.g. metastasis? -Yeah, so we haven't looked
into that part so far, whether — what are the effect
on the metastasis. It — There has been some work that has gone
into the degradation, how the extracellular matrix,
if it degrades, how it affects the metastasis,
but specifically for our case, we have not looked
at how it affects. So we have been planning
some studies where we are going to create
a metastatic model by injecting tumor cells
in mice and then see if
the D-cell biofactory — how effective
will that be. However, like, starting from
the first principles, I think, like, that is going to be
very similar to what, like, our hematological
malignancies are, and so if they are —
If the tumors are metastasizing, D-cells should be able
to find them.

We have not yet
tested though. -If neither of you have
anything you want to add on to that,
the next question was for Josh. "Burns are serious problems leading to enrichment
of nonfunctional circuits due to intrinsic evolution. Could you share more
perspectives on this issue, how to minimize mutations
or cope with them?" -Yeah.
Thanks, Nicole. I think this is
the first from Tae-Suk. This is a good question. So one of the nice things
about this kind of circuitry in mammalian cells is that this
is far less of a problem than you might expect
compared to other systems, like bacterial cells
or mammalian cells when you're trying to do
something like biomanufacturing, cranking out
recombinant antibodies.

You definitely see
those selections. The nice data I showed you about the functional performance
of these receptors actually all happens
in an expression regime where we can't detect
any evidence of burden from just a couple of parts. We've started now in our systems a rigorous kind of monitoring
for when burden kicks in and have an actual quantitation
pipeline in place there, so it seems like burden
is a real thing, but you can actually,
with a relatively — If you're doing things
that involve a relatively small
copy number of genes, and that's a level of expression
that's totally adequate for doing all the processing
that I mentioned before, you can sort of avoid burden.

There are also strategies
for managing burden. So far, we've been able to get
pretty far with simply avoiding it. -All right. Tara, "For mass production
of platelets, can this be done entirely
with chemical induction and sorting,
quality control, or do you also need flow
to generate active and compositionally homogeneous
thrombocytes? -You know, this is
a great question, so my answer
is sort of twofold. The chemical induction
that we use is actually to produce large
numbers of the megakaryocytes because what we have found is that when using
media conditions, we get a very
heterogeneous population, but we're able to improve
that population to get our megakaryocytes, so that's where our
chemical induction takes place, but then the question
regarding the flow generated to make these thrombocytes
or the platelets, yeah, so we use a device
that does include flow, a filtering system, and we then
get our nonactivated platelets, and what we've been working on
is actually isolating. We don't actually run them
through flow again, or we have certainly done
sorting with the flow cytometer, but we are concerned
about activation by doing that
because it's such a — We have to use
a different sized nozzle to keep them
from getting activated, but we're very interested in the
production of high throughput, and so we want to avoid the sort of sorting
from the flow cytometer, and so we're actually
designing a second part to our production
of platelets so that we can actually purify
them, the nonactivated ones, so then we can then go get a
little bit closer to the clinic.

-Caroline, did you
have a question? All right. Parijat, "Have you tried
the biofactory genetic circuit in other cells?" -Yeah, actually we have
implemented the circuit with NK-92 cell lines
type of an NK cell, and we observed that
the biofactory is active, same as in D-cells,
primary circuit cells, primary D-cells
and in NK-92 as well. Yes, we have.

-Josh, "What is the sensitivity of the hypoxia sensor
to oxygen levels? In other words,
can it respond to low oxygen or to particular level
of oxygen?" -Yeah, this is
a great question. In fact, this is what motivated
us to look at these circuits. In reality, there are
intermediate levels of hypoxia between normoxia, physoxia, low hypoxia,
intermediate, very low hypoxia, and so a lot of initial work was done at the extremes
of this regime, and we started doing
some characterizations at those levels that are more what you might find
near the vascular law, the periphery of a tumor,
things like that, so there is a — The answer is,
there's a gradient.

There seems to be a gradient. It's not entirely clear
how that maps to a single cell level decision, like whether
what we're looking at is that there are
a high frequency of cells converting when you get at those
intermediate levels there, but that's a big part
of what we're doing right now with those kind of comparisons going on between
the native machinery, various versions of sensors
for the native machinery, sensors that have augmented genetic circuits
of different types, and part of what the Synbio
investigation is letting us do actually is figure
a better answer to your question because there isn't one
from the literature that was compelling to us, but we're using these synthetic
tools as interrogations as well as
for useful functionalities that it's enabling us
to build a mathematical model and hopefully have
an answer for you because I don't know
the answer right now, other than it's graded.

Yeah. -Another question for Tara,
"With your engineered platelets, are you concerned about them
becoming spontaneously activated and not delivering
to the right locations?" -Yeah, that's a good question. So we are actually
making triggers in the engineer platelets, and so we're using
a directed evolution approach to build receptors so that we're actually
disconnecting the receptors to bind to proteins,
pro-clotting proteins, and actually
re-engineering them to bind to either
tissue-specific proteins or also do more of a systemic
pharmaceutic induction. -Okay, thank you. If there's no other questions, I think this is right on time
for the next session.

-Good afternoon, everyone. Welcome to our Showcase:
Part Two. My name is Jiahe Li. I'm Assistant Professor in
Bioengineering Department from Northeastern
University. I hope everyone has enjoyed
the talks since yesterday. We have heard a lot of wonderful talks in engineered
and synthetic cell as in microbiome
synthetic biology field. I'm sure everyone in
the audience are also anxious about learning
about some exit events and technology
using bacteria, engineered bacteria
using synthetic biology, so the notion
is that the bacteria, especially commensal
bacteria and probiotics, they are kind of in an ecosystem
with our host, so if engineering
a cell is to target whole cells
as a therapeutic manufactory. Engineered bacteria,
on the other hand, is trying to, you know,
repurpose those bacteria and harness the crosstalk
between bacteria and hosts towards a therapeutic
application. So in the next 45 minutes, we're going to have
three wonderful speakers, going to talk about their
research on engineered bacteria for therapeutic intervention, and we'll save our Q
and A question at the end, so please leave your question
in the chat if you have a question
during the talk.

And without further ado,
I will have Andy take away. -Okay.
I assume folks can hear me. Thank you very much
for the invitation. After that last session,
I'm just stunned because of the amazing complexity
of eukaryotic circuits, and so in some ways
what I'm going to tell you is just really, really simple. It's an attempt to essentially
provide a treatment for Parkinson's by not administering levodopa
as a chemical compound, but rather having bacteria
produce levodopa inside a mouse, and we would not be able
to take this on were it not
for a great collaboration with Anumantha Kanthasamy, who is now at
the University of Georgia. When we started this,
he was at Iowa State but has recently moved, and I'll show you some of
the joint work along the way, but initially, I'm going to talk
about the development of parts for, again, some relatively
simple circuits, and so what we want to do is be able to regulate
the production of DOPA based upon
the sensing of DOPA, and to that end, Ross Thyer,
who is now a professor at Rice, identified a DOPA-responsive
transcription factor.

It's actually called DopA
or PP2551 and did directed evolution of it
using a platform that our lab calls compartmentalized
partnered replication, and actually, this is
a really nice platform in which you regulate
the production of a thermostable DNA
polymerase in individual cells, and library members that best
upregulate the production of the DNA polymerase
can then be self-selected via emulsion-based
amplification, so if you're
a good transcription factor, you make more
DNA polymerase, and then in the context
of thermocycling and emulsion, you amplify yourself. So using this technology,
Ross was able to, as you see there
along the bottom, go from a relatively
unresponsive wild-type gene to a variety of variants, including the so-called
mftE variant that gives very,
very good activation with very low background, so like Josh before us,
you know, we think that we can make some
really good sensor elements that can be used
for further engineering. We immediately used
this particular sensor element to improve the production
of DOPA itself. Now, there are a variety of ways
in which you can add that additional hydroxyl to make
the catechol that is dopamine, but we tend to use
the HpaBC system, which is originally present
in E.

Coli, but we used once again
our CPR system to now improve the function
of those enzymes because now that we had a sensor
that had a high dynamic range, we could look for the production
of more DOPA and again craft
a synthetic genetic circuit that could be used
for directed evolution, so that circuit is not
so much shown here, although KOD,
the DNA polymerase, was used during its evolution. Instead, I'm showing you that
when we use this circuit now to make DOPA, and we detect the DOPA with the
engineered transcription factor by producing GFP —
So it is slightly complex — We see once again,
we go from wild-type production by the wild-type enzyme
that is very, very small to very, very large increases in the production
of this amino acid, and again, it's in part because of the introduction
of the sensor, which then led to
the improvement of the enzymes, which now feed back in terms
of being able to sense the improved production.

I just want to point out that
as an aside, this work has not
only proven very useful in terms of engineering
a microbiome, as I'll talk about
in just a moment, but DOPA itself is
a pretty interesting compound, and so in collaboration
with the Army, shown down there in the lower
right, Jimmy Gollihar, we've been able to
make polydopa, or melanin, at very, very high levels, up to and exceeding
1 gram per liter, and we're beginning
to pilot the production of interesting
melanin-based materials using these same constructs. However, what we're
going to do now is show that we can make DOPA
from bacteria for implantation
in the mouse microbiome.

Now, initially, once again,
we're just going to measure output of DOPA
and from HpaB and HpaC, using the engineered
transcription factor, and this was work
by Shaunak Kar, where along
the bottom there, you see that as we induce
the circuitry, which is under the control
of TEDR with tetracycline we produce
increasingly more DOPA, which is sensed by
the transcription factor, and then one can see
that you have, at higher levels of induction,
much more DOPA produced. Now, this was all done in the E.
coli Nissle strain, the probiotic strain that can be used potentially
for colonization of the gut, but these are just basic example
of how we can do that.

And now I'm going to shift over
to Dr. Kanthasamy's work. We handed off the basic
constructs to him, and he was able
to reproduce now using electrochemical
detection of DOPA, rather than our transcription
factor-based detection, the improved production of DOPA
by our constructs. Now, I should point out that
this particular construct does not yet use
the improved enzymes. These are still
the wild-type enzymes, and I think a reasonable
question during discussion is, "Wait, are you going to produce
a black film in the stomach?" for example,
and the answer is, "We don't know yet,"
but presumably, you know, because DOPA is
a natural amino acid, we can tune production
so that we're not getting films, and the environment
of the stomach is very different than the environment
of flask anyway, but nonetheless,
we can produce DOPA that can be seen now
in culture media other than by using
our transcription factor, but more importantly,
when those bacteria are now given via gavage to a mouse,
they will establish themselves for a few days
following inoculation, and during those few days
following inoculation, we see an increase in DOPA
production in the bloodstream, and not only in the bloodstream
but importantly in both — in striatal samples
both of DOPA and dopamine, where the DOPA
would need to get to in order to potentially lead
to improvements in dopamine-deficient
neuron function in Parkinson's disease.

Now, these results were
initially in C57 black mice and wild-type mice,
but what we've also done — What, I should say,
Dr. Kanthasamy has also done is show this
in the so-called MitoPark mouse. Now, this is a model mouse
system for Parkinson's. It does involve
dopamine-deficient neurons created in an interesting way that involves engineering
of mitochondrial factors, but up there in the left,
what you can see are these sort of
locomotor activity assays basically watching
where a mouse goes, so there's where a mouse goes
when it's a control mouse, a C57 black mouse. There in the MitoPark mouse,
in the middle is where a normal, untreated MitoPark
mouse would go, and it has reduced
locomotor function, but now look at the right there
where upon the administration of these
DOPA-producing bacteria, at least early on, what you can see is improvement
in locomotor function. Now, is this surprising? Probably not because, of course, Levodopa would produce
the same thing, but this is microbially
produced DOPA in the context
of the mouse itself, and over there
on the upper right, we can see that the DOPA is once
again showing up in the striata and where it should be
in order to actually effect
this sort of improvement.

So where do we go from here? We don't just want to sort
of produce DOPA constitutively because obviously bacteria
may rise and fall in the context of the microbiome
or over time may be diluted, and so we've made what we call
a homeostasis circuit. It's basically a gene circuit that is meant to
keep up production of DOPA. Now, it's under T7 RNA
polymerase control. That wasn't the case for any
of the previous constructs. -Two-minute warning.
-Thank you. What we have, essentially,
is auto-induction of T7 and auto-repression
of T7 via — and so we have sort of measured
both positive and negative feedback loops.

We've also put this on a plasma that can be transferred
between strains, and right now we're not going to
look at DOPA production. We're just going to look
at a GFP output, but essentially in E. coli Nissle, the strain
operates as it should. Depending on different
strengths or promoters, it produces a constant
amount of GFP over time, and it can be modeled,
and we basically can mathematically model
it not only in E. coli, but also in
salmonella typhimurium because we're interested potentially and
in the longer-term having a variety
of microbial platforms, which we can use to introduce
DOPA-producing strains into the mice to produce DOPA
under different regimes, whether it's homeostatic, whether it's in
a pulse generator, whether it's in a sinusoidal
wave, what have you. We basically can do sort
of regulated production and have that regulation
be under the control of the organism itself,
irrespective of that organism. So conclusions: I hopefully
have shown you that when you use some good directed
evolution of parts up-front, you can make some
pretty simple circuits that nonetheless have
very good performance.

DOPA overproduction can enable
both biomaterials production and microbiome-mediated
therapies. Modular construction
of circuitry can support a variety
of regulatory paradigms, and we're only beginning
to explore those, where really we're going
to have a transfer function from circuit-in
to mouse health-out, and that's exciting
for me because again, even though the circuits
are simple, they're effecting a complex
physiological function. In the future, we want
to determine how different
controlled release modalities will impact the MitoPark
mouse model, as I just said, and I really want to explore
the notion of essentially inoculating via
horizontal transfer, not just putting the bacteria
in via gavage, but putting the DNA in and seeing to what extend
we can get it established and self-regulated
at the microbiome level. And with that, I'm through,
and thank you so much.

I think I talked about the
important people along the way, and so this is the final slide.
-All right. Yeah, I want to thank
the organizers for the opportunity
to give a short talk on some of the work we're doing,
thinking about engineering and predicting
the human gut microbiome. So the human gut microbiome
is a major determinant of human physiology, and many of these interactions
between the host and the microbiome are mediated
through chemical transformations that can either be beneficial,
such as buterate production, or, you'll deleterious
or kind of, you know, disease-promoting,
such as TMA production, which is linked
to cardiovascular disease, so we're really
interested in thinking about these
chemical transformations that are mediated
by these microbial interactions, so the interactions
between the different species in the community
have a big impact on this complex
metabolic profile, metabolite profile
that's produced and degraded by these communities, and ultimately we want
to understand the interactions between the different organisms
in this community so that we can actually
control the microbiome by exploiting the ecology, but also thinking about
new mechanistic control knobs that allow us to sense different
states of the microbiome and make specific changes
in the environment to shift microbiomes towards
a beneficial state.

So fecal microbiota
transplantation has been shown to be highly effective
in some cases against C. difficile infection, and one
of the challenges with FMT is just the unknown factors
including pathogens that can be transferred
into patients. And so I think there's a need
for thinking about these
next-generation therapeutics where we can identify lower-order
reduced-complexity consortia that are highly well —
highly characterized, and we can predict
their behaviors, but also combining that
with engineered microorganisms that have circuits to sense
and respond to the environment. And we can start moving
towards a more well-standardized,
programmable, predictable and having very type-controlled
specificity of the functions. Our lab leverages
a bottom-up approach for the design of synthetic
microbial communities where we have a toolbox
of different microbes, and we can study them
in a variety of different ways in the lab ranging from
kind of lower-throughput, more precise control to, you know, millions of different
measurements simultaneously using
droplet microfluidics.

So one of the biggest challenges
when we start thinking about bugs as drugs,
as, you know, a therapy is kind of embracing
the complexity of the system. So just given the presence
and absence of different species in a synthetic microbial
community, there are — It grows exponentially
with the number of species, so it's kind of
a combinatorial explosion, and so we need to think
about ways of navigating this really immense
design space, this functional design space
of synthetic communities. For engineered species,
we need to understand what sensors can be built,
what actuators we can use, exploiting molecular
mechanisms of interaction and also selecting chassis
with high enough fitness that the organism can stably
persist in a microbiome for a period of time to perform
its function in a robust way. So we've been thinking
about the design of synthetic
microbial communities, focusing on the health, relevant
and beneficial metabolite, butyrate, which is produced
by specialized microorganisms in the gut microbiome, and the ecology
of these organisms is really important
to understand in order to design communities that can enhance the level
of this beneficial metabolite.

So our goals were to develop
the capability to predict butyrate
concentration by understanding the interactions
that are occurring in these synthetic communities and by building communities
from the bottom up. And this project involved
using computational modeling and a design-test-learn cycle with a 25-member synthetic human
gut microbiome community where we could identify
the significant interactions that were taking place
between butyrate producers and the other members of the
community and design communities that actually could enhance
butyrate production, and this, again, is a really
large functional design space. Just considering all communities that have all five butyrate
producers in our system, there are over 1 million
possible communities. So on the left is what we call
a functional landscape, where every single community,
every single point, is a different consortia
that we could build in the lab.

And using this iterative
design-test-learn cycle, we built a predictive model that could forecast
butyrate production across a wide range
of different communities and also identify molecular
mechanisms of interaction that are taking place,
and these represent molecular control knobs
for starting to think about how to manipulate
butyrate production. So one of the interesting
findings from this study was that we identified consortia that substantially increase
butyrate production far beyond what the mixture of five
butyrate producers can do alone, and this is due
to microbial interactions, interplay of resource
competition and pH and other molecular mechanisms that are taking place
in these communities, just highlighting the need
to understand interactions to enhance
beneficial functions produced by these
synthetic communities. So I'm going to transition
to talking a little bit about how we can identify
new mechanistic control knobs within these systems,
and ultimately we want to build these smart, you know,
next-generation therapeutics that can sense the environment,
that can make, you know, predictable change and robust
change in the environment. And so we need to understand
the contributions of individual metabolic pathways and to understanding how
to manipulate these systems.

And we've been focusing on
a very highly abundant commensal species
called Bacteroides uniformis, which is one of the most
highly abundant microbes in the human gut microbiome, and so we sought to engineer
the uniformis' function. So this organism
is high-fitness, and so we think about introducing new functions
into this organism so the function will be more
stable and effective in the gut. And we started by looking at
polysaccharide utilization loci, which are complex
metabolic pathways that degrade dietary glycans
and host-derived glycans and produce molecules that
influence the ecology of the gut and also have a big impact
on human health. And to do this, we developed
a CRISPR tool for Bacteroides that allowed us to make
these markerless deletions in Bacteroides uniformis
and constructed a library of 23 different
polysaccharide utilization loci mutants to understand
the contribution of these pathways to fitness.

And we barcoded each strain
so we could pool them together and look at their
combined fitness in a variety
of different environments, both in vitro and in vivo. So using this library, we identified the functions
of several PULs in this system and their interactions
with different dietary glycans. We also found that PULs do not
always increase fitness, and this is evidenced
by a mouse experiment we did where combined
all of these barcoded mutants and introduced them
into a germ-free mouse and looked at their abundance as a function of time
on different diets that vary
in the glycan composition and also the fat content. And we found that deletion
of specific PULs such as PUL 37 here
in kind of blue color substantially increased
the fitness of Bacteroides uniformis
in the mammalian gut. And actually
most of the PULs — Not most, but some of the PULS
demonstrated much higher fitness than the wild type. And so this shows
that there is a cost to the presence
of these PULs in environments when the pathways do not
have a specific benefit, and this has to do with
the nutrient environment.

So the nutrient land —
-Two-minute warning. -Thanks. And so we identified
these two PULs, which have a huge impact on the fitness of uniformis
within the gut. The last thing I'll mention
is understanding how these polysaccharide
utilization loci influence butyrate production. Because polysaccharide
utilization loci degrade these chemically
complex glycans, they can influence
community interactions, especially interactions with
butyrate-producing bacteria. And so we assembled a set
of synthetic communities to study how the contribution
of individual PULs within B. uniformis influences
interspecies interactions in a glycan-dependent manner but also in
a butyrate-producer-specific manner, too,
because we have five different butyrate
producers in our system. And we find that these
polysaccharide utilization loci have a huge impact
on butyrate production, and so they can be targeted
as control knobs for influencing the ecology and also
the beneficial functions produced by
these gut communities.

So, in summary, you know, synthetic communities
can be used to understand the mechanisms
of community-level functions. We have models that can predict
and design these communities and ultimately optimize
the therapeutic potential of different
microbial communities, and we want to target
microbial interactions and new mechanistic
control knobs that allow us to manipulate
beneficial functions produced by these microorganisms
in the gut. So — and I'll thank my lab. I did mention several of the — Dr. Jun Feng
and also Dr.

Ryan Clark, who led these projects. And I'm happy to take
any questions at the break. -Great, so I'm going
to be talking about engineering bacteria
for cancer therapy. I'll just give you kind of
an overview of projects going on in the lab
and talk a little bit about some of the projects
funded by NIBIB specifically. So many people in this audience
may be familiar with the concept that using bacteria as a cancer
therapy is quite an old concept. It was observed over 150
years ago that, serendipitously, bacteria could have an effect
on the growth of tumors and reduce their growth. And it wasn't until
several decades later that William Coley really
pioneered this at a large scale, injecting patients with bacteria
to try to treat their cancers. And for this talk, I think the
only point that I wanted to make is really there was many species
of bacteria that were used, but some species,
like streptococci and cerecea had significant successes in
bone and soft tissue sarcomas, more than 56 percent
overall response rate, and these didn't work
that well in carcinomas. So for this work, William Coley
is known as the father of immunotherapy because the bacteria was thought
to stimulate the immune system by a yet-undiscovered mechanism,
and at that time, really, using bacteria was considered
largely unsafe because of the availability
of antibiotics and the beginnings
of radiation chemotherapy, and so the fields
really died off.

But today, really, synthetic
biologists start thinking about, how can we engineer
safer strains of bacteria while still maintaining
and improving upon some of the efficacy? So one of the things
that Coley didn't quite appreciate at that time, and it was discovered
several decades later, is that certain
bacteria species can actually selectively grow
inside of tumor environments, and this is a prototypical
experiment shown here where you intravenously
inject bacteria.

It could be systemic delivery
by other mechanisms. And the bacteria land
in healthy tissue in the mouse, and they land in tumor tissue, but they're clear out
of the healthy tissue, and they grow inside
of the tumor tissue, and that happens specifically
because the core of the tumor where these bacteria actually
survive is immunoprivileged. There's not a lot of immune
surveillance, and they use it essentially as
a safe haven to grow and thrive. So here, you can see in the
mouse these bioluminescent E. coli are are localized
to the hind flank tumors, and they produce
this nice biodistribution with high concentrations
of bacteria in the tumor and none found
in the healthy organs. So the limitations
of the approach is that bacteria
can target tumors about 1/2 millimeter
to 1 millimeter cubed, so post-angiogenic-switch, you need to really have
this necrotic core that has no immune surveillance. And like I mentioned,
I just want to emphasize it's a selective
amplification mechanism. It's not a targeting mechanism,
per se. It did land
in the healthy tissue. It would land in the tumors,
but at some time point later, after 24 hours, they're cleared
out of the healthy tissue.

So certain bacteria species have been shown to grow
in these tumors selectively. They have to be able to survive
the hypoxic environment, and it's also lower-pH
than healthy tissue, and so salmonella typhimurium and E. coli are typically
the most widely used strains
that are used because they are the most
genetically contractible. And probably what I think
is that, you know, the most exciting aspect
of this approach is that bacteria
have the ability to colonize essentially all types
of solid tumor, so they don't really care
so much about the genetics. It's not a particular receptor that they're targeting
or something about the cancer cell pathway
that we're trying to target but rather the hallmarks
of the solid tumor, which is a very universal
targeting mechanism. So in our lab and in other labs,
we've shown that bacteria robustly colonize
many different mouse models, both human and mouse tumors
in subcue, orthotopic, metastatic
and also genetic. There are neurotoxinous
models of cancer, so it seems
to be quite robust. I'm not going to get into it
today, but there's actually
some data in humans demonstrating colonization with
different species of bacteria.

So it's fairly well-established
that bacteria can colonize tumors,
but just the bacteria alone do not have a very strong
therapeutic effect, and so the focus has really
shifted towards thinking of, how can we engineer
these bacteria to locally produce therapeutics
to augment their efficacy. So the approach that we take is what we call
intratumoral bioreactors. These bacteria
live extracellularly within
the tumor environment, and they can make
peptides and proteins and other types
of therapeutics. There are other groups
that are working on bacteria that invade cancer cells and deliver something
intracellularly, just want to make
that sort of distinction. So the advantage here is that
we can make high concentrations
of drugs locally. These could be drugs that have
already been developed or thought about,
but we can actually make them with essentially a better safety
profile and delivery vehicle.

We can reverse and kill
these bacteria and tumors. We've shown that. And then there's a lot of
synthetic biology approaches that we can use
to continuously sense and respond
to the tumor environment. So we think about these
four axes in our lab, manipulating the strains
of bacteria, the types of payloads
that they produce, the circuits that control
their behaviors and also their clinical use, so I'll just give you
a couple of snippets here. And one of the first things
that we did is work with
a synchronized lysis circuit that was originally developed by
Omar Din and Jeff Hasty's lab, and it consists of a positive-
feedback quorum-sensing circuit where bacteria
reach a critical mass, and then they produce
this phage lysis chain that allows them to first open and then release effectively
a recombinant payload as well as all of their
intercellular contents and the LPS of the bacteria that can act
as an adjuvant response.

And then these bacteria, not all
of them lyse as a population. Some regrow, and basically
we perform this lysis in a tunable manner. So we went on to explore
what kinds of therapeutics might be best for this, and we looked at CD47,
the CD47 alpha axis, which exists
as a "don't eat me" signal on the surface of cancer cells,
also on that surface of essentially
every healthy cell in your body, and we wanted to block this
interaction using a nanobody. And so when we expressed that
in bacteria, intratumorally injected it into
tumors in an A20 lymphoma model, we saw complete clearance
of these tumors, and then an abscopal setting where we just injected
a single tumor, we saw a response
in the untreated tumor, as well. So we went on
to investigate this, and we found
that it wasn't the bacteria trafficking from one tumor
to the other.

There is no bacteria found
in the untreated tumor, and there's no therapeutic
really translocating to the other tumor, as well. The bacteria stay relegated
to the treated tumor, and at the very end,
the tumor is essentially gone, so there's no bacteria there, and there's none
in the healthy organs. But rather than the bacteria,
we're stimulating an intratumoral
T cell response, and we see that
in the treated tumor as well as the untreated tumor. Here, you see increased
proliferation activation signals that interfere on gamma, and we actually showed
this was tumor-antigen specific. So if you take that tumor,
the untreated one, and you do
an ex-vivo restimulation with the MHC class-one
restructive peptide, that corresponds to that unique
A20 E cell, lymphoma cell, we can actually see a response in just the CD8
T cell population, as we expect. So we did a lot of controls
showing that the antibody, miap301,
blocking CD47 has some efficacy in coinjection with bacteria,
the lysate of the bacteria, the recombinant antibody,
et cetera.

They all have some
therapeutic effect, but you really only see
this sort of striking clearance on the tumors with the live
and lysing bacteria that can continuously deliver
the therapeutic over time. And so the sort of commutative
mechanism that we have now, I didn't really get
too deep into this, is that the CD47 nanobody
blocks the interaction that leads to more phytocytosis, and bacteriolysis also
helps increase phytocytosis, so you have these
two modes of broadening and increasing
antigen presentation that leads to this durable and antigen-specific
T cell response. So we've more recently made
some technological improvements, put this into a probiotic E.
coli Nissle strain, expressed some checkpoint
inhibitor nanobodies and shown that,
with a single injection, we could get clearance
of these tumors and outperform the antibodies
at their highest concentrations.

And then in colorectal
tumor models, these checkpoint inhibitor
nanobodies don't work that well, and so if you express
a cytokine like GMCSF, you could improve
the therapeutic response. So it worked —
-Two-minute warning. -Trying to figure out,
thanks, what the ideal combination
of therapeutics to make is. And so going back to that
William Coley story that I told you, certain strains
of bacteria work for sarcomas, and other strains, like we see
BCG for bladder cancer, we're trying to really think
about how we can scale this and search different strains
in a high-throughput fashion. So this is a system that Tetsu
built in our lab called bacteria spheroids. You can grow these fluorescent
bacteria into these tumor spheroids
by adding the antibiotics and applying
a couple of lush steps, and this mimics the 3D tumor
environment inside of tumors.

If you don't have antibiotics, they just sort of grow
everywhere. We can do some neat imaging, tracking of the bacteria
intensity over time for at least a week or two, and I'm just going to skip
through this quickly. We can do characterization
of genetic circuits. We can clone different kinds
of therapeutics like toxins, anticancer peptides,
and look at the spheroid growth, and probably what we'd never do
in animal models is test all the different circuits
across all the therapies.

And here, you can see
we've got quorum-sensing versus quorum-sensing
plus lysis. Some therapeutics
are improved — are not improved by lysis,
and some are improved by lysis, and that helps us understand what are the kind of
release mechanisms we need. We've built a variety
of different biosensors, hypoxia,
lactic and pH, a couple of them together
and a couple of them to the essential gene
production,and this improved some of their performance
in biocontainment for these specific signatures. Sorry, I'm not going to go
into this too in-depth. And then lastly we've been
also thinking about cancer immunotherapy
more broadly, and checkpoint inhibitors, CAR
T cells and oncolytic viruses are already making an impact
in cancer immunotherapy, and we think bacteria
have sort of a unique niche because they can call
on these colonized — these solid tumors in a nice,
antigen-independent fashion, so there's a lot of unmet
needs in cancer that could be addressed
with the use of bacteria.

And so we're testing
combinations and also comparisons across
these different technologies. The most recent one,
we call ProCARS, or probiotically guided
CAR-T cells, where we have bacteria
colonize the tumor cores and produce these
synthetic antigens that the CAR-T cells could be
engineered to recognize, and so this gets around
some of the limitations of CAR-T cells with finding
a safe and specific antigen or antigen loss
escape mechanisms, and we've shown
some specific lysis in vitro as well as some combination
and enhanced effects in vivo.

-Time has expired. -Okay, perfect, so with that,
this is just an overview, again, of a couple of different axes
that we've been working on, and here is a list
of all the people who worked on these projects
and some of our funding awards. Thanks. -Let's give our three speakers,
Tal, Andy and Ophelia,
a round of applause. Thank you so much for
your wonderful presentations. We're going to have a Q and A
session for the next 15 minutes, and then we're going to have
a break, I think still at 2 p.m. After the break, we will hear
about the vision and perspective on FDA,
particularly regarding engineered mammalian cell
and bacterias. So I look at the questions
by the audience. I kind of wanted just
to summarize of some actually
pretty common to each view. For example, I'll start off
with the first questions about biocontainment. So, for example, you know,
maybe a question — Maybe, Andrew, would you want
to kind of share your opinion about the biocontainment? And do you — What do you think
about biocontainment in terms of your system
for DOPA production in vivo? -All right.

So I think it's going to depend
upon what the functionality is. Biocontainment or, you know, sort of adding something back
is different than going — going to be different
than biocontainment for something
being taken away. So I think that for Tal and I, they're going to be different
answers to that question. My own answer is that I don't
see a real issue, per se. I see an issue
with misregulation, and there were
some nice questions during in the chat about, you know, how do you
really fine-tune this? And that is really important, but just in terms
of containment, since it's like a drug
being added back, I see that
as less of an issue. But now, I turn it over
to Ophelia and Tal to further comment. -Yes, Ophelia, do you want
to kind of add something? So perhaps your consortia,
for example, if you engineer consortia
somehow outcompete the population
of indigenous bacteria, would you expect any,
you know, failproof approaches to make sure you are containing
your engineered bacteria within the right dosage? -Yeah, I mean, I think
this is an interesting question.

If you coupled some, you know,
nutrient that is uniquely used
by the consortia, or, say, the consortia kind of
depends on this nutrient, and then remove
that nutrient, then, you know, and that's supplied
through the diet, that could be one mechanism
of containment, biocontainment, is just the dependency
on something external, like, some, you know, dietary
input, so that's one strategy. I mean, other strategies
involve, you know, like, killswitches
and other mechanisms, I think those
negative selections can sometimes lead
to escaped mutations, so if you had
something that was — had a kind of fitness burden
or fitness was toxic, that might lead to evolutionary
escape on a faster time scale.

-Yes, so for Tal,
I think, Ophelia, Andy, there were a couple of involved over the years, I guess, with
prior systemic delivery, right. If you want to target breast
cancer with immunotherapy, probably with your system, may be required a little more
exchange in the control over timing and spatial delivery
into the tumor to enhance
the immune therapy, would you commenting
on this? What kind of potential
biocontainment strategies in your system
for immunotherapy? -Sure, yeah, so I was thinking
about biocontainment was maybe coming
from a different aspect, which is, there have been now
many phase-one trials and a phase-two trial
going on now with the bacteria
cancer therapies. And one of those trials
is done intratumorally, so not systemically. In fact, I think that
there's even maybe less control needed
in that case, and one of the methods
that they — that this is some logic
that has sort of utilized is to make these strains lack the production
of essential gene.

And they can get some of it
from the environment, and they can get
some of it initially when they were injected,
but they could only grow in the tumor environment
for a couple of days, you know,
the maximum of a week, and then they're
actually cleared. But you sort of rely on the
maybe more practical aspect for cancer therapy, which is that you could
do injections every 2 weeks, and that's not just the case
for intratumoral injections. That could be the case for
systemic injections, as well, so I think in some ways, for us,
it comes down to a matter of, how much efficacy
do you really need? If you don't need that much,
and you don't need that much bacteria, you can weaken
the bacteria significantly and just rely on sort of
containment by oxytrophy and things like that
and then, you know, just do multiple injections
over time or rely on antibiotics to clear those bacteria
in the tumor niche or something like that. -Wonderful, thank you. I think I see some, you know, very technical questions
for Andy.

How will you be able to control
the level of DOPA? I think those question were
raised before we talked about, you know, the switch
in homeostasis of, so would you
be able to elaborate on the DOPA production? -Yeah, so I think I did
a poor job towards the termina
of my talk explaining what those points were were
actually different T7 promoters that led to constant
homeostatic production. So those were things
we could fix. We didn't have to dial them in
with ATC or anything else, and so in that way,
we really can finely regulate a fixed amount of production,
at least of GFP. We'll eventually see with DOPA, and the additional,
you know, advantage with DOPA is we will have
the transcription factor that can also serve
as part of the negative feedback loop, if we wish. So I think we will be able to do
very, very fine tuning, and that's sort of what we were
showing there at the end. Now, in vivo, of course,
is a different story because there as — Again, there were
questions in the aside. It's going to be how many
bacteria, you know.

What is going to be the PKPD,
per se, of both those bacteria
and of the compound made? And so those are just things we're going to have to
work out over time. We have the tools to do it, but a lot of those are things
where I think it's just going to have
to be approached empirically, and I think that's probably true
for my copanelists, as well. -Thank you.
Yeah, maybe one correction before I ask to Tal and Ophelia. So I think the case,
Nico is asking Andy, you know, is your DOPA circuit
based on plasmids or integrated
into a chromosome? Guys, I'm also interested
in this question, too, because we're actually
building some circuit into plasmid
versus a chromosome, and we have some stability issue
or cloning difficulty in terms of strong promoter —
the strong promoter making cloning
super difficult.

That would be able
to kind of provide a little bit more detail
about this here, your circuits, whether it's based on
the plasmids or chromosome. -And so just real quickly,
and I really do want to hand it off to the others, you know, there are ways
to improve plasmid stability. There are ways to decrease
fitness effects including of something
that is almost always a killer, T7RA polymerase. Anyone who has used it knows
that that thing is just a beast. But, you know, as Ophelia
was also saying, you can include things
like toxin-antitoxin systems if you really want
to keep something around.

And as Tal was saying,
maybe that's not a problem if you're injecting
every 2 weeks, or you're giving gavage
every 2 weeks, right? There's an advantage
to it goes away. -Yeah, thank you. Yeah, just to let you know
that I heard from Dave that we don't have a time limit
at 1:45, so we can continue past it if you have more questions
and if you want to answer. So I'm not just limiting
the question to Andy, so I guess for Ophelia,
for butyrate, I think Casey can ask
the same question. How would you be able to control the level of the butyrate
in vivo? I guess you can through
the direct control or the composition
of consortia, so what kind of design criteria
or strategy have you implemented
to control the butyrate level to make sure it's not
too high, not too low? -Right. I think that's a good question,
and there are some — There's some evidence that
butyrate at very high levels could potentially be toxic and
cause some deleterious effects, so we may not want to just
maximize butyrate all the time but turn it on, say, when the host has inflammation
or if there is some dysbiosis.

And I think the simplest way
to think about this is mainly in a single organism, you know, we are working
on butyrate pathway engineering and commensal strains that allow us
some synthetic control at the circuit level
of the production of butyrate, so that's a very straightforward
kind of idea where you would have, you know,
some sensor for inflammation and then turn on butyrate
production when it's needed, and that's something that
we're working on in the lab. In terms of a consortia
dynamics, we've been working on using,
like, recurrent neural networks, which is a deep learning method,
to design the dynamic behaviors of synthetic microbial
community metabolite production so we can actually
identify communities that have desired dynamics. And so one potential dynamic
input could be, you know, turning on for
a period of time in the host, and then butyrate
kind of goes away.

And so, you know, we've tested
these communities mostly in vitro,
but perhaps in the future one could think about
designing dynamic functions of these synthetic
microbial communities, and of course,
we would need to understand how the host contacts
in the mammalian gut influences
those dynamics, so I think it's kind of
a longer-term vision. -Thank you, and I will
get back to Ophelia. I thought I saw that Nico
just posted a question, but I was going to ask Tal about his engineered E. coli
probiotics. So they would learn about,
you know, how healthy were engineered T cell
for cancer immunotherapies, but you are doing something
different with the engineered E. coli to somehow interface
with the immune cell, which would be able to call
on some strengths and advantages
bacteria system over, you know, mammalian cells
that other people are kind of working
on such as T cells circuits.

-Yeah, I think the biggest
advantage in bacteria even compared
to oncolytic viruses is really that they target
these tumors in what I keep, you know, referring to as
the antigen-independent fashion. So they really don't seem
to care that much about the genetics of the tumor
but rather the that there is a solid tumor necrotic core
that they could colonize. So that makes it
extremely robust. I think, you know,
90 percent of patients are the ones
with solid tumors that are — That's where the largest
unmet needs are, so, you know,
the challenges with CAR is, I think everybody knows there's
the safe and specific antigens and antigen escape
and all those kinds of things, and with oncolytic viruses
it's also antigen, their targeting-based approach,
same as oncolytic — same as nanoparticles.

So I think this kind of
universal concept as a colonization by bacteria,
it could be in a tumor. It could also be, I don't know,
in other niches as well, could act as sort of, like,
a first-stage colonization that then interacts
with other living medicines or other technologies. I think we're starting to see
people build these interacting communities
as medicine. Sometimes they're microbe,
microphilias where — But I think microbe cell and microbe virus
and microbe particle, those kinds of things, I think,
are starting to be developed to try to address the challenges
of each technology. -That's — Yeah.
I guess for my own experience, I worked with mammalian cell
and bacteria. I think that the process of learning bacteria is much
shorter than in mammalian cells.

Yeah. And let's move to the question
from Nicole. I think Nicole just posed
a question for Ophelia. "For designing synthetic
consortia, does your math focus
on bacterial composition or on the functional microbiome? Is the goal to engraft complete
microbiome in a human host?" -Yeah, I think, you know,
function is, we think,
more important than composition. There's a lot of
functional redundancy within these communities,
and so we — Our models
do predict composition, but I think function is what
we're really trying to understand
how to design, so, like, the metabolites
that have these links, these mechanistic links
to human health and how we can
manipulate them, and, you know, in some ways
predicting function can be easier
than predicting composition, one, because
we have a limited — We have a fewer number
of outputs for our model versus, you know, say 25 species
in our synthetic gut community versus four metabolites
that we're trying to predict, so we actually can achieve
pretty high accuracy of prediction
of those metabolites.

So, yeah, we're thinking
a lot about function. I think in terms of engrafting
complete microbiomes, I think maybe
our longer-term vision would be to have communities
that are really stable and perform functions,
but they also are able to sense and respond
to their environment, and so we can design these,
like, support communities to enhance the functions
of specialized organisms within the community,
and so in that case, we would be engrafting
a reduced-complexity microbiome that could persist over
a desired period of time to perform
a beneficial function, and we need
to definitely think about the biocontainment
question, especially for
the engineered organisms.

-Thank you. I think Kelsey just asked about, "What about the speed
of sensing and actuating?" I guess this question maybe,
Andy, do you want? Or Kelsey was asking
about overall questions. Maybe I'll clarify —
Go ahead. -Yeah, I mean,
this is sort of a derivative of what I think of
as the PK/PD question. A little bit of the answer is
going to have to be empirical. I actually think that
for most bacterial-based sensing at least,
it's going to be on a time frame that's very fast relative
to physiological change. Now that's not going to be
true necessarily if we start doing
neural engineering, another topic entirely, but I think for a lot
of the metabolic engineering, we're probably okay. -All right.
I think — Thank you so much
for answering the question.

We have more questions
that we can answer during the time limit. I just kind of remind
the audience that we will hear about FDA on their perspectives
on the translation of either
the mammalian cell or microbes. All right. Thank you, everyone,
for your attention. -So welcome back to our final
session in the 2021 NIH Syn Bio Consortium meeting. Joining us next we have
Karen Elkins and Carolyn Yong from the FDA,
and they're going to give us a little FDA 101
as we finish our discussion about translating synthetic
biology products to the market. Karen is the Associate
Director for Science in the Office
of the Center Director, and she is responsible
for the strategy and coordination of
scientific research activities conducted by CBER
including development of a research
and scientific agenda that supports the development
and evaluation of CBER-related products. Her expertise is in infectious
disease immunology where she researches
the nature and immunity against intercellular pathogens and develops correlates
of vaccine-induced protection for vaccines
against intercellular pathogens, and Carolyn is the Associate
Director for Policy in the Office of Tissues
and Advanced Therapies, and she manages
guidance document and policy
development for OTAT and liaises with external
stakeholder organizations to advance OTAT-regulated
product development and regulatory review.

Her expertise is in cell therapies and biologic device
combination products, which bridges CBER
to CDRH and CDER, so I think we got a tremendous
amount of power and knowledge at the FDA right here
for us to enjoy, so I will turn it over
to both of you. Thank you so much
for being here. -Thank you, David,
and thank you for including us in this really fun meeting. I have learned a lot,
and I am a fish out of water because as David alluded to, my background is
as an immunologist, and I have regulated vaccines
for a long time, but fortunately Carolyn knows
everything there is to know about the cell,
gene and tissue therapy world much closer to the heart
of this audience, so she and I are going to switch
off on the course of this. We are available to you for
questions and after the fact, and we will leave you
with some resources. And so we're going to try
to keep this light, and I hope it's not too light. We know it's Friday afternoon, and there could be
nothing more deadly than listening to a regulator on a Friday afternoon
after a long meeting, so we're going to fly
through some topics and try to give you a flavor
of what you might need to know.

So we're from the FDA,
and we're here to help you out. You're probably most familiar
with our role as regulators of new product, but you may not know that
we also have a large research and testing program, that we inspect
production facilities, and we have a large
communications effort, and we are divided into a number
of organizational units. The three that probably matter
most to you all are the Center
for Biologics, the Center for Devices
and the Center for Drugs. The products that go to
each of these centers are sent there
by a combination of history and legal definitions
of what belongs there.

There are in the current era
some jurisdictional overlap, and we often discuss
who gets what. The authority for our
regulatory activities comes from a long
history of legislation, but the two foundational ones
are the Food, Drug and Cosmetic Act,
which goes back to 1938, and the Public Health
Service Act, which is relatively more recent. The point of this legislation
is what you might not expect, which is that products
are regulated according to
interstate commerce. The original definition was
that a product must be approved and licensed
in order to cross state lines. Now you might think that you
could go across the street and test your product
because of that, but the interpretation of that
legislation over the years has evolved into any component
of the product that needs a license
in order to cross state lines, and given the global
and international sources of components
as a practical matter, you need either a license
to shift and market something, or you need an exemption
to that requirement, which is an investigational
application, and —
focus on today.

If your product comes
to the Center for Biologics, it will go to one of our
three product offices, which are in teal here. The most likely one is probably
the Office of Tissues and Advanced Therapies. The products that we regulate
are loosely captured here. They range from allergenics,
blood products, vaccines, phage and live biotherapeutic
treatments, xenotransplantations and some devices
that are directly tied to the biologic side, but most of your interests
are probably in OTAT, which is where Carolyn is from. I don't expect you to read
this giant list except to say that
it is a giant list, and it grows all the time
thanks to you all and encompasses gene therapies,
stem cell products, products derived
from mature cells, differentiated cells, blood
and plasma-derived products and combinations
of all of those, which are certainly
not uncommon.

So what we're going to talk
about a little bit today is the life of a product
and how you get it made and into clinical development
and some of the programs that we have to
expedite developments, particularly for things
that are very unique and address unmet needs and as was alluded
to this morning, your opportunities to interact
with us and to get our guidance and input into
your product development plan, and as I mentioned, we'll
leave you with some resources. This is one of those
standards cartoons that FDA-ers always use about
how things move from basic science,
which are — And they're there
a known amount of time into some translational phase, into an actual
clinical development phase and then out into the market, and the time frames
that we listed on here are also the wild variation of
5 years plus or minus 50 years. I would point out that
this timeline has been blasted by the development
of COVID vaccines as well as a panoply of other
COVID treatment products and gone from years down to
months or in some cases weeks.

So there are — We usually divide regulatory
discussions into several phases. At the start of entry
into human beings, you would need to submit an Investigational New
Drug Application, known in the lingo
as an IND. In the Center for Drugs, these are called
New Drug Applications or NDAs, but it's the same idea. It's an exemption to that
premarketing interstate commerce requirement, and that allows you
to go into human beings. If things go well, then you
submit a request for a license, a biological
license application, and if things go very well,
and that is approved, and the product goes out
into the market and into doctors' offices, there are many activities
after approval, and those include
inspections, all lots are released
formally. Usually the license comes with
some postapproval commitments like long-term safety follow-up
or changes in the product, and those all result
in amendments to the license. So what do you do if you think
you've got a candidate or you're getting close? I think we can debate the value
of this slide, and it's the world
according to Karen for sure, but my main advice is to seek
experienced collaborators.

Find yourself somebody who is
a regulatory affairs professional who is steeped
in the knowledge of FDA options
and interactions, an organization such
as contract manufacturer that has the capability,
knowledge and experience to produce your product under
good manufacturing practices, which I'll say a bit more
about in a moment, and find a group that has
serious clinical trial experience and experience
with proper statistical analysis of the resulting clinical data,
so I know that's a little scary because that probably means
multimillion dollars' worth of consultants
and collaborators right there. I do that to emphasize that
prompt development in the US and in the world today is really
not for the fainthearted. These are now highly
specialized activities that do take specialized
knowledge and experience. There are good, bad
and indifferent groups that do these activities
out there. I encourage you to seriously
consider finding the good ones. -I guess while we're waiting
for Karen, I can just say a few things
about manufacturing or what we in being referred
to as chemistry, manufacturing
and controls' test do with manufacture
of the products and all the controls in place
to make sure that a product can be used in the clinic,
you know, is produced, and here on the slide
you see a reference to good manufacturing
practice experience now.

Those are in the regulations,
so by the time, you know,
a product reaches Phase 3 and going through submitting
for a license application, FDA would expect that, you know,
you have full GMP in place. There's a little bit
more flexibility when it comes to
early-stage clinical trials. You know, sometimes you're still
working out your manufacturing processes, but I can tell you that
in the rare disease space, you know, where there's
smaller patient populations, and clinical trials
may go a little bit faster than those for more prevalent
conditions or diseases, it's really important to lock
down your manufacturing in your early stages because, you know,
sometimes we experience that it's not necessarily
the clinical data that's holding up
product development.

It's actually locking down
that manufacturing process and getting those GMPs in place
so just an important, you know, aspect of getting
regulatory consultants involved and gaining that
manufacturing experience. -Assays, at all points
in the production process, you will need to test and document the tests
for all areas, and that includes
sampling during production and then having
a battery of tests that will assess the critical
attributes of the product to be used for lot released
and later for stability, and in order to select
and develop those assays, product characterization
is critical, and this obviously is
highly product-specific. Some things that are commonly
used to assess product
quality include aspects related to its physical nature
like the pH, the osmolarity. You may want it to be aggregated
or not aggregated and all the things
that it's composed of.

We insist that you have have
a test related to identity to determine,
to demonstrate quite clearly that what you think is
in the vial or the syringe is what's in the vial
or the syringe but also to distinguish it
from other things that are made
in the same facility, and if you can't imagine
how that could be an issue, I give you Emergent, AstraZeneca and J & J vaccines
earlier last year. Assays for purity
are necessary to demonstrate either its bioburden
or its sterility as well as
its physical purity as well as things that
reflect safety such as endotoxin or the potential
to cause allergies or to remove toxic raw materials
in the course of production. Probably the single
most important test, and this can often be
a group of tests as well are something
that measures potency, meaning the assay that
demonstrates a product capacity to affect a given result
so for example, if it's critical
that a gene be expressed in order for the product
to be active, then an assay that demonstrates
the gene is in fact expressed is important.

It's also important to develop in parallel assays
of clinical material which are expected
to be important during the clinical
development phase. Next slide, please. Also in parallel with
product data are data that support
your clinical testing plan. For example, you should have
some justification for putting your stuff
into humans in the first place, and not just any humans
but humans that are relevant
to the target disease. You will need a rationale
and data to support the starting
clinical dose as well as the route
and regiment by which it's administered and to establish feasibility
and reasonable safety of the product
administration procedure itself, which is a separate idea from
the dose or the route, per se.

The clinical trial will no doubt
propose certain patients, and so the data to support
the patient eligibility criteria is important and based
on the nature of product, to identify and suggest
potential toxicities to guide clinical monitoring and
physiological determinations. Next, please. So when you've got all of that,
come see us, and we will — When you bundle up
your IND application, we will be looking
for a scientific rationale that supports moving forward, a detailed description
of the product and its manufacturing and all
that quality-control testing.

The litany is safety, identity,
purity and potency, but these are not four tests. These are our battery of tests that address many angles
of the product. Typically we're interested
in pharmacology and toxicology testing and then
the clinical trial protocol with all
the associated approvals and a robust statistical
analysis plan. Next, please, and that whole package
of information goes to a team. It goes to a team
that's usually headed by a regulatory product manager
who puts it all together, a pharm-tox reviewer,
a clinical reviewer who is responsible
for the clinical trial design and then for monitoring
the outcomes as the clinical
trial progresses, a stats reviewer
and then a CMC reviewer, which is somebody like me
who is responsible for understanding
the scientific rationale, the product itself
and its production and probably for
the clinical assays, in vaccine's case,
immunogenicity and others, whatever is applicable.

Next, please. That group of people
will render a verdict on whether things
can progress and then will be with
the product as it moves forward. The goal of the clinical
development plan is to generate data that supports moving
further into people and potentially out
into the market, and so the entire
clinical development plan should be designed
to get there. There are expedited
development programs and incentives for serious
diseases with unmet needs. There — We will be looking
throughout the phases at product quality
and consistency. Often forgotten is the need
to demonstrate consistency of manufacturing. Making it once to go into
your Phase 3 trial is wonderful, but you need to be able to make
it again and again and again so that when it's out
into the big world, it performs in the same way, and that you are confident that
it'll perform in the same way.

The clinical development plan
should be designed, obviously, to generate evidence
of effectiveness and an acceptable safety profile
in the population of interest and that it can be delivered. Often also overlooked is that
companion diagnostics or devices that are necessary
to deliver the product need to be codeveloped
in real time, and all of that should move
forward to demonstrate that the product's benefits
outweigh its risks forever. Next, please. I'm sure you already head
about the famous Phase 3 trial, Phase 1, 2
and 3 trial scheme.

1 starts in small numbers
and is focused on safety. 2 moves into larger numbers
and continues to focus on safety and then starts to get an idea
of dose and responses, and Phase 3 trials are designed
to build on the 1 and 2 steps to really get to
evaluating safety and efficacy in larger numbers. Now, the numbers can range
from 10 or 50 if it's a small, rare disease
population to 30,000 if it's a COVID vaccine.

Next, please. All of the product
life cycle should be accompanied
by a focus on safety, but as things progress
through clinical development, product characterization
comes first and should guide
the development of assays that need to be
qualified and validated, and those should progress as the
clinical development progresses, and the all-important
potency assay may start out
as something rudimentary but needs to progress
to its full-blown final state by the time
you get to Phase 3. In this way, we hope that safety
and quality are designed into the product
from the get-go. Next slide, please.
Yeah. Everybody always wants to know
about how and what to look for
in product characterization.

Unfortunately, neither of us
will give you a set answer to that
because both the expectations and the nature
of the characterization are incredibly
product-dependent. They depend on things
like scale, which could be something
very small to something huge. They depend on the nature
of the manufacturing procedure, and one particular aspect that is probably of interest
to this group is, if something is very short-lived or available
in very short supply, doing all that comprehensive
product characterization and developing the assays accordingly can be
a real challenge. What we are looking for
throughout that is where the risk is and the level of attention
that gets focused on, ever any particular aspect should be focused
on the level of risk. Next slide, please. Everybody always wants to know, does my stuff
have to be made under GMPs in order to start
my clinical development? The most easy answer to that is,
no, not necessarily, but GMPs are not a one thing.

It's not just your product. It's all of the parts
of your production and the assays
that go into lot release and quality-control assays. Things can start out
in a softer place, but we do expect that, that will
be progressed as time goes on. Full compliance is required by
the time you get to licensure, and it's very much
to your advantage to have everything
buttoned down and locked up by the time
you enter into a Phase 3 trial. If you are lucky, things move
fast, so getting to the — all the GMP-related aspects
early and often is really highly advisable, and we certainly don't want
product considerations to hold off
a useful therapy.

Next, please. So things work, hallelujah. Now what? Now you bundle it all up
and send it to us, and we reanalyze it it
over and over again. We don't believe
the press releases. We look at all
the clinical trial data. We look at every aspect
of the manufacturing, and we reanalyze
it ourselves, and we come to a consensus
on whether — what the state
of the outcomes really are. Stability needs to be
demonstrated. The stuff needs to last as long
as you intend to use it in the clinic. Some lots are tested as CBER. That's not so common anymore,
but it does happen, but for sure a lot-released
testing plan needs to be in place. Another time-consuming
activity is labeling, meaning not just the tag that
goes on the vial or the carton but the package insert that describes all of
the clinical data that is most relevant
to the product approval. Next slide, please. So that's the whirlwind overview with a few
technical interruptions of the life of a product, so I'm going to switch
over to Carolyn now and let her tell you about
how to interact with us and what we can
do for you.

Carolyn, take it away. -Good afternoon, everyone. Looks like we're in
the home stretch, so I'll be going over
the different mechanisms available
for interaction with FDA and some special programs
available through our sponsors. You know, we can't stress enough
at FDA to community with us early and often, you know,
when the timing is right, so the information I'll be
presenting is super-specific, but some of it
is cost-cutting and may be applicable
to products regulated in, for example,
the Center for Drugs. Now developers of biological
products have multiple opportunities
to interact with FDA, as you can see
on this slide. These interactions may include both informal
and formal meetings where the formal meetings follow
these established processes and timelines
for requesting, scheduling
and documenting such meetings. These processes and timelines
are defined in the FDA regulations,
guidance documents and SOPP. Of if you have informal
interactions like teleconferences or e-mail
exchanges that someone can request outside of or
in addition to formal meetings.

But today I'll primarily
be focused on meetings prior to submission of an IND,
which I think is the information that will be most helpful
to this group. So when developing a novel
product, the sponsor may want to obtain
CBER's advice on the data needed to support
the submission of an IND. Now the INTERACT
and pre-IND meeting may be held at these early stages
of development, here defined as preclinical
stages in light green. Next slide, please. So the INTERACT or an Initial
Targeted Engagement for Regulatory Advice
on CBER products is an informal,
nonbinding meeting. It's an opportunity for sponsors
developing novel therapies to request feedback
at an early stage of development and was put in place
to encourage this early interaction
of sponsors and replace what used to be
the pre-pre-IND meeting process across the Center
regarding preclinical, manufacturing and clinical
development plans, so you might ask, what's
appropriate for an INTERACT? This is intended to discuss
the development of, you know,
what we would consider innovative investigation
of products that has unique challenges
such as unknown safety profiles, complex manufacturing
technologies, the incorporation of innovative
medical devices and use of cutting-edge
testing methodology.

So the appropriate timing
for an INTERACT is when a sponsor has
identified a specific product to be evaluated
in a clinic study and has conducted
some preliminary preclinical proof-of-concept
studies with that product but hasn't yet designed and conducted definitive
toxicology studies. So considerations for whether
the status of product development is premature
or too advanced for an INTERACT meeting further discussed
on CBER's web page, and, for example,
a request may be premature if, you know,
you haven't identified a specific
investigational product.

You haven't provided preclinical
proof of concept or other pilot data particularly with
the product of interest, so additional details on development program
qualifications for INTERACTs, how to request one,
where to send it is available on the SOPP
at the bottom of the this slide. Next slide, please. So certain meetings held over
the course of product development as shown
on the earlier slides have been assigned
to be either type A, B or C, and FDA has defined timelines
for these different categories of what we consider
formal meetings, which are explained
in FDA guidance.

So there are type A, B and C meetings after submission
of an IND, but as I said, we're going to
focus on early-stage meetings. So now I'll talk about this
type B, pre-IND meetings. Pre-IND meetings can be valuable
in planning a biological product
development program, especially if questions aren't
fully answered by available guidance documents
or other resources. They also provide information
that will assist sponsors in preparing to submit
a complete IND application, which is important
because we want to reduce the risk of setting up
and not being able to proceed, which we call a clinical hold. The primary purpose of this
meeting is to review and discuss the design
of animal studies needed to initiate human studies
as well as, you know, the design of the
initial clinical IND studies, also the best approach
for formatting and presenting your information
in your application to facilitate FDA review.

Again, sponsors are unsure
whether an interact or a pre-IND meeting
for the current state of development is appropriate. So in general a pre-IND meeting
is appropriate when you need to define
the manufacturing processes. You've developed some assays
and preliminary lot release criteria
for your product, and you've completed
both proof of concept and possibly some preliminary
preclinical safety studies and want to move on
to more definitive studies. Next slide, please. So now, let's switch gears
and talk about some expedited programs
available at FDA to, you know, what I imagine
would be applicable to many subtypes of products in
development by this group here. So all five expedited programs
listed here represent efforts to address an unmet medical
need in treatment of a serious condition
or disease.

Accelerated approval,
that pathway has been used primarily in settings where the
disease course is really long, and an expended period
of time would be required to really measure the intended
clinical benefit of a drug. An application for a drug
will receive priority review designation
if product, if approved, would provide
a significant improvement in safety
or of effectiveness. And fast track designation
is intended to facilitate development
and expedite review so that a product
can reach market faster. The last two,
breakthrough therapy designation and regenerative medicine advanced therapy
or RMAT designation are two of these
expedited programs that, again, I think will generally apply
to products described today. So breakthrough for a product
which preliminary clinical evidence
indicates that the product demonstrates
substantial improvement over what's already
available out there.

And then a product eligible
for RMAT designation may be for
regenerative medicine therapy, which includes cell
and gene therapy, therapeutic-
tissue-engineered products and human cell
and tissue products. So for those products
developed under breakthrough and RMAT, sponsors receive
much more intensive guidance on efficient drug development with increased interactions
and communications with CBER. Next slide, please. So fast track,
breakthrough therapy and RMAT designation
are the same programs with different funding. Sponsors can apply for
and receive more than one designation
for a given product. In the next slide are tables comparing the criteria
and benefits for each program. And I'm not going to go
into too much detail, as those can be found
in guidance listed at the bottom of the slide. But we thought it would be
really important for relatively new
or inexperienced product developers to at least be aware that these programs
are available and we kind of have
to state early on that you intend to apply
for these designations.

And, you know,
it should be noted that the level of evidence
required for breakthrough is actually higher
for fast track designation. Next slide, please. So again, this is just
a continuation of a comparison table,
highlighting those benefits that those programs for sponsors for the treatment
of serious conditions. Just to highlight one,
advances of RMAT designation include actually all
the benefits of fast track and breakthrough
designation programs, including those early
interactions with FDA and, in fact, may be eligible
for priority review. Next slide, please. So what if you aren't developing a specific
investigational product but rather have an advanced
or platform technology and canvas for the development
of such products. CBER has the CBER
Advanced Technologies Team or CAT program, which was established
to promote dialog and education and input amongst CBER staff and between CBER
and prospective developers, innovators of advanced
manufacturing technologies at meets
focused on novel technologies that can have a significant
impact in this field on product development, manufacturing,
control strategies.

They may also have
regulatory implications, so it's good to give us
a heads-up on these things. Manufacturing
and analytical methods, you know, for products for which
CBER has limited experience with the manufacturing
and development process are definitely welcome
to come in and discuss. Next slide, please. So last but not least, FDA has
numerous guidance documents published that describe FDA's
interpretation of our policy on a regulatory issue. So these guidances usually
discuss a more specific product or issues
that relate to the design, manufacturing and testing
of regulated products. They also relate to
the processing and contents of an IND for a specific product
and inspection and enforcement discretion
on enforcement policies. So it's important to be aware
that guidance documents are not the regulations. They're not binding
and of course any alternative approaches
may be chosen to comply, you know, with
regulatory requirements.

So here I've listed just some
specific guidance on things that may be applicable
to some of the products introduced during this meeting. This is just a flavor
of what is available, so we really encourage you
to search on CBER's web page for other relevant guidance. Again, the interactions with FDA
such as the pre-IND meetings can be really valuable
in planning your product
development program, especially if your questions
aren't fully addressed with these guidance documents. So, again, it's also important
to remember that FDA expectations
are highly product-dependent. Next slide, please. On this slide, I've just listed
a guidance available to authors to facilitate navigating
the interaction processes. Next slide, please. And then you can refer to the
following two guidance documents for more information
on the designation criteria for the expedited programs
that I described earlier.

And with that, I'll leave you
with Dr. Elkins and my contact information,
as well as some other resources that may be of assistance to you
during your product development. And again, I'd like to thank you and the organizers
for the opportunity to participate
in this fantastic meeting. And we really look forward
to seeing you come in and discuss your products
and plans with FDA. Thanks. -Okay, thank you, Carolyn. Thank you Karen. Yeah, so we've got a little time
now for a little Q and A before we close out the meeting.

So welcome all
and any questions here. As you're gathering
your thoughts I've got a couple questions. And so speaking very generally,
what would you say would be the most common pitfall
or missing detail when researchers engage
the FDA for the first time? Like, what do you wish people
had for the most part that didn't have the first time
they came to you? What's missing? -Great question.
I think the — When somebody
brings me a vaccine, and this may be less applicable for less stable
or long-term products, my overarching question
is always going to be "What's in the vial?" If you're going to take
the vial, the syringe or whatever applies
and stick it into a person, I want to know what's in
the vial, soup to nuts. And I think people don't have
an understanding of what that question
means in the first place and what it takes
to answer the question in the comprehensive nature
of what we're interested in. But in the clinical development
sense, physicians and all of us care about what we're about
to inject into people.

So trying to really think
big picture about what you might inject in people,
including what you expect to and what you don't expect to,
is, you know, sort of — I want people to think about
that scientifically as well as logistically and practically. Carolyn, what's your answer
to that one? -I think for these novel
products, product characterization. I think it's similar along
the lines of what's in the vial. Want to know what
the clinical product is then. Although you don't need to know
the exact mechanism of action or actions of your product, it is important to know
what your product is and, you know,
we often use the term identity it's one of those requirements
that you really need to satisfy in order for us to be
more confident about the safety and effectiveness
of the clinical product.

-Cool. I also wanted to talk
about acceleration for a few minutes too because
everybody wants it now, right? Now is better than later. So COVID accelerated
the process, of course. And it sounds like there are
other expedited avenues for certain types of products. Are there any general
considerations about how a faster process
could be more common as opposed
to being the exception? -Carolyn, you want to take
the first stab at that? -Right, you know. I went over these expedited
programs intended for, you know, the serious
conditions or diseases.

There are criteria,
and, you know, we've experienced — We understand, you know,
that unmet need. So, but I guess, you know,
you're saying if you have an indication that, you know,
may not meet those criteria, how you would go about,
you know, maybe fast-tracking
that process. You know,
it's so product-specific and the burden
of the clinical evidence, you know,
you have clinical trials. You have confirmatory
clinical trials. There's some of that maybe close
to post-approval but really is — I have to say,
it's part of the process. -So you mentioned
the advanced — What was the word again,
advanced technologies? -Yeah, technologies?
-Yeah. –
The Advanced Technologies Team? -It sounds like a lot
of platform technology might kind of fall
under that umbrella in general. And I think a lot of synthetic
biologists think about themselves, you know, as toolmakers
and kind of platform makers that have
a lot of broad utility. So under that advanced
technologies avenue, are things — Do you notice
a different failure rate going through
the advanced technologies path versus maybe, you know,
a more traditional path? Maybe it's not
a useful question.


So the advanced technologies — -I don't know.
-It's not a pathway. FDA regulates medical products. We don't license, approve,
clear technologies, per se, especially
platform technologies. But because they're becoming
such an integral part in all of these — the development
of novel therapies, it's just really important
that we understand those manufacturing processes
because, in my mind, that falls under the chemistry
manufacturing and controls so that we know that you're
producing the product consistently
and, you know, quality that can be administered
to a human.

-Cool.
-David, I feel complied — compelled to go back to your
question about acceleration, so everybody wants faster,
better, you know. The counterbalance
to that question, and this will sound like
a really wonky regulator, I'm sure, is that there really aren't any
shortcuts for a lot of things. If you bypass it now,
it will bite you later. And a lot — You know, I think there's
a sort of a sense out there in some corners of the world that FDAers kind of
lay awake nights thinking up new laws
and regulations and ideas to throw sand
in the gears and just for the heck of it.

Almost all of our
legislative history, never mind our practical
regulatory expectations, are driven by a disaster. Something went wrong. It's a rather reactive
kind of thing. Something major went wrong, and we tried to learn
from that experience and incorporate
the knowledge from that
into our future expectations. So it's not that we're laying
awake nights trying to think of roadblocks. It's more that
we've seen some — a problem arise and are trying
to help all of us address that problem before it becomes
a giant problem. So even for the current example, I think you've seen it
in the media about COVID vaccines,
they're under EUA, there's a large segment
of the population that thinks that that's awful
because that means corners have been cut
and things haven't been done. That's actually not true and
because of the nature of these. On the other hand,
the gap before the EUA and the final authorization is to finish every
last assay validation, every last reference standard, every last inspection
of all the many plants brought online and so forth. But the guts of the expectations
were all up front at the EUA time.

And I think that kind of idea
and example is applicable to all the products
under your discussion. You know,
so while there are ways, we are all interested
in getting things in to people who need them
as expeditiously as possible, at the end of the day
everybody needs to be satisfied that they're well produced,
well controlled, and that we understand the risk benefit relationship
very thoroughly. -Yeah.
-On top of that, you know, there are regulations — I mean, the products
have to meet the regulations. And, you know, as FDA gains
experience, you know, the burden
of the data required, you know, it —
as we all know, you know, a better understanding
of what we're dealing with can certainly help development
of potency assays and all these things that people
actually pinpoint in terms of what might stall
product development. Again, you know, there's just
no getting around regulations.

There has to be purity, potency,
identity. So it's an understanding
and experience. -Yeah, and our purity, potency,
safety, identity litany came
from an understanding of what can go wrong
and blow up in your face if you don't have
those bases well covered. -No, yeah. I do appreciate you taking
the time to elaborate on that because I think it's something
we all probably know but it's always good
to hear it again. So thanks for taking your time.
-That is a wonky regulator Friday
afternoon thing to throw in. -We'll have a fair chat later. I have a question in the chat
here for Karen. -Oh, I should — -"What's the guidance
for the length of time for following
clinical trial participants involved in gene
therapy studies? Does this change depending
on the phase of study?" -That's actually better
directed to Carolyn. Your turn. I can tell you about vaccines. -So we actually have a guidance
on, you know, long-term follow-up for subjects
that receive gene therapies.

We could always provide
that link. But there is specific guidance
that talks about it and even breaks it down
a little bit in terms of the type
of gene therapy products. –
A lot of those kind of questions you can reframe for yourself from the point of view
of "What are you worried about? What could go wrong?
What's the — And that goes back
to the science base and your knowledge of how you're
expecting the product to work, what you know about the nature
of the condition being treated, and so forth. But the decisions are not cut
and dried even by class of product. They're very tailored to the,
"What are we worried about? What could go wrong? What could we miss?
What could we not understand?" That's from the patient
safety point of view.

In the clinical trial level
it's, "What can we learn
across the population?" -Great, so I think we'll — I'll just slide right into
closing out this meeting here. I'm just going to take a few
more minutes to thank you, Carolyn and Karen. Appreciate you giving us
a little bit of your time to give us
a little foundation for, as hopefully everyone
is thinking about, how to engage the FDA. It always seems like
a monumental step, but I think you really laid out
a nice framework, so thank you. -It's what we're here for.
-Anyway, yeah, you know, in just the last couple
of minutes here, I just wanted
to kind of just — a global thank-you to everyone,
you know, from all the feds that came out to support this
at the NIH and the FDA, and then, of course, the entire
synthetic biology community that, you know,
came to participate, ask insightful,
awesome questions and tell us about
their wonderful innovations coming out of their labs. So I'm really glad you could all
join us for these 2 half-days.

It looked like
a pretty good turnout. And so I'm excited,
enthusiastic about that. And so I hope what that
translates into is another great turnout
hopefully a year from now and hopefully in person because,
you know, I got to look through all the little black boxes
with people's names on them, and I wish I was able
to welcome you all in person. And so I'm hoping that the 2022
NIH SynBio Consortium Meeting will be held on NIH's campus
in the Natcher building. We've already got the space
reserved, so mark your calendars. It's going to be November 7th
and 8th next year.

And then hopefully
in the ensuing months or so, down on the website
at the bottom of the screen, we will include more information
as it comes online, so yeah. So yeah, I just really hope everyone got a little something
out of this meeting. You take away something,
a little bit about translating something from your lab
into the FDA one day. So, yeah, I guess with that,
I'm just checking the chat box. Thanks, Tae Seok, yes,
hopefully in person. And so I think with that,
we'll close out the meeting. Thank you once again to everyone
who was a part of this. Thanks, again..

You May Also Like

2021 NIH Synthetic Biology Consortium Meeting – Day Two – Ai Biotechnology