University of Southern California Dissertations and Theses

Rapid microfluidic prototyping for biological applications via stereolithographic 3-D printing: a journey

(USC Thesis Other)

Rapid microfluidic prototyping for biological applications via stereolithographic 3-D printing: a journey

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Rapid Microuidic Prototyping for Biological Applications
via Stereolithographic 3-D Printing: A Journey
by
Kenmond Hall Pang
A Thesis Presented to the
FACUL TY OF THE USC Viterbi School of Engineering
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
MASTER OF SCIENCE
CHEMICAL ENGINEERING
December 2020
Acknowledgements
I would like to thank the amazing support structure that I have had during my graduate studies here are
the University of Southern California.
T o my advisor, Dr. Noah Malmstadt, thank you for all your amazing support and encouragement you
have given during this trying 2 years at USC, especially during the COVID-19 pandemic.
T o my collaborators, W an-Zhen (Sophie) Lin and William Evenson, thank you for bringing me onto this
project. I don’t know what I would have done for my thesis otherwise, but I do know that meeting you
two and our friendships are some of the things that I will cherish for the rest of my life.
T o my committee members, Dr. Richard Roberts and Dr. Malancha Gupta, thank you for taking the
time out of your busy schedules to support my endeavors to complete a master’s thesis.
T o my research group members, nothing beats seeing Majed or Matt past midnight in the lab with me,
or having hours long topical discussions with Sepehr, Lucia, and Ahmed. I also can’t forget my deskmate
Zikai just for coming along the ride with me. All of you have have helped to support me mentally in my
times of crisis. I will miss you all.
T o the members of the Roberts research group, thank you all for just being amazing people. You are all
so knowledgeable and willing to help me whenever I needed it. Thank you Chris, Golnaz, and Kaori for
all your help and care during these 2 years.
T o my friends that were with me thick and thin through this entire process. There are too many of you
to list, but all of you deserve credit for me being able to survive.
Last but not least, I want to extend the greatest of thanks to my mother and sister. They will always be
my bedrock, supporting me in whatever I do. As a family, you can’t ask for any better.
ii
Contents
Acknowledgements ii
List of Tables iv
List of Figures v
Abstract vi
1 Background 1
1.1 Stereolithographic Printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Asiga Max X27 UV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Microuidics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Navier Stokes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Reynolds Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 mRNA Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Building the Automated System 9
2.1 Fluid Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3 Microuidic O-Rate Selections 13
3.1 MFED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Selection Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Challenges 21
4.1 Protein Sticking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.1 T ranscription T ests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.2 Fluorescent Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Resin Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2.1 Recipe Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.2 Resin T esting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5 Current Projects 29
6 Conclusions 30
7 Future Directions 31
References 32
List of Tables
1 Advantages and challenges of microuidics . . . . . . . . . . . . . . . . . . . . . . . 4
2 T able of dissociation rate constants k
o
. . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Resin comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
iv
List of Figures
1 Image of the Asiga DLP printer model Max X27 UV . . . . . . . . . . . . . . . . . . 3
2 Schematic of mRNA display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3 Automated mRNA display example . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4 Example IR sensor placement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5 IR sensor apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6 IR sensor operating schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
7 Schematic of how an MFED operates . . . . . . . . . . . . . . . . . . . . . . . . . . 13
8 Picture of an MFED deconstructed . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
9 Picture of an MFED with blue liquid owing through it as a reference . . . . . . . . . 14
10 Binding vs. owrate selection experiments . . . . . . . . . . . . . . . . . . . . . . . 16
11 MFED washing experiments and theoretical enrichment . . . . . . . . . . . . . . . . 18
13 MFED vs. manual selection gel images . . . . . . . . . . . . . . . . . . . . . . . . . 20
14 Incubator CAD drawing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
15 Transcription tests in the incubator with low yield . . . . . . . . . . . . . . . . . . . 21
16 Design of experiment for transcription . . . . . . . . . . . . . . . . . . . . . . . . . 22
17 Transcription protein deactivation conrmation . . . . . . . . . . . . . . . . . . . . 23
18 Fluorescent microscopy - negative control . . . . . . . . . . . . . . . . . . . . . . . . 24
19 Fluorescent microscopy - positive sample . . . . . . . . . . . . . . . . . . . . . . . . 25
20 Molecular structure of PEG-DA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
21 Molecular structure of IRG-819 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
22 Molecular structure of ITX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
23 CAD drawing of resin height testing apparatus . . . . . . . . . . . . . . . . . . . . . 27
24 Schematic of how a prolometer operates . . . . . . . . . . . . . . . . . . . . . . . . 27
25 Layer height vs. Light Source Energy graph for resins . . . . . . . . . . . . . . . . . . 28
26 Non-porous silica beads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
v
Abstract
Potential targets for new therapeutics and diagnostic tools are being discovered at a greater rate
than existing methods are able to determine ligands that can bind to these targets[1, 2]. New methodolo-
gies and devices need to be developed to be able to match the pace in which targets are discovered and
one of those pathways is via stereolithographic printing[3, 4, 5, 6]. We have used stereolithographic 3-D
resin printing to rapidly prototype bio-compatible microuidic devices to facilitate an automated sys-
tem for mRNA display. Automating and developing microuidic analogs of mRNA display process will
drastically reduce the sample size, process cost, and total time needed to process a target[7]. One of the
most successful devices we developed and stereolithographically produced, is the microuidic enrichment
device (MFED) which enables kinetic o-rate selection of ligands without needing an exogenous com-
petitor. Simple and known kinetic equations are also able to model ligand binding and dissociation in the
MFED, with experimental rate constants agreeing with other independently measured constants. After
MFED variables were tuned such that ligands bind and dissociate to a selected target based on anity,
we demonstrated that the MFED can enrich a ligand at least 4 times greater than that of a standard man-
ual selection. While developing the MFED, there were challenges in the production of bio-compatible
microuidic devices due to protein sticking (adsorption) to device surfaces, specically the inner chan-
nel walls. This was discovered by using simple design of experiment and conrmed by uorescent mi-
croscopy. T o address this problem, new bio-compatible resins were developed and experimented with,
to determine their ecacy and suitability for the mRNA display automation project. The overall goal of
the project to automate mRNA display is a lofty goal that is still unaccomplished, but with interesting
challenges and side projects solved, we are step by step attempting to complete that goal.
vi
1 Background
I was brought in by PhD student Wan-Zhen (Sophie) Lin, of Dr. Noah Malmstadt’s group, at the
end of December in 2018, to help with a new project in collaboration with Dr. Richard Robert’s group,
and his PhD student William Evenson. The original project proposal and overall goal of the project is
to create an automated system that can perform a cycle of mRNA display, with minimal inputs and zero
human supervision. T o accomplish this task, two kinds of technology were essential in the fabrication
and implementation of a system to automate mRNA display: stereolithographic 3-D printing and mi-
crouidics.
1.1 Stereolithographic Printing
Stereolithographic printing has been researched as early as the 1970’s, with the rst patents for
the foundational processes issued in the 1980’s. Stereolithography, as the word is dened, can be broken
down to its two roots: stereo and lithography. Stereo is dened as relating to solid forms having three di-
mensions and lithography as a printing method that reproduces an image from a standard onto a printed
medium. This describes the stereolithographic process that has developed today to become commonly
known as stereolithographic apparatus (SLA) printing. SLA printers work on that basis that a 3-D model
or object can be fabricated by a photochemical process that uses the energy within a certain wavelength of
light to trigger the polymerization and cross-linking a solution of monomers and oligomers. This photo-
chemical reaction happens layer by layer, per an example on an X-Y plane, with each printed, contributing
to a Z-plane, until a full 3-D model is printed. The light source used to trigger the cross-linking process
is the dening feature that separates the dierent types of stereolithographic printers. All stereolitho-
graphic printers are technically SLA printers, but when generally referring to SLA printers, the common
perception is of a single point laser as the light source activator. SLA printers draw the layer image pixel
by pixel using the laser light source, before moving on to the next layer. The second type of stereolitho-
graphic printer is a digital light processing (DLP) printer. Though a DLP printer is still technically a type
of SLA, as it is also an apparatus that prints via stereolithographic principles, the way it diers from a
1
typical SLA printer is that DLP printers use a projector to selectively photo-activate pixels when printing
a specic layer. Even though there are developing and more advanced SLA printing technologies that
have been discovered, SLA and DLP printers are the most commonly used type of printers available, and
currently the most commercially viable. Commercial stereolithographic printers can now cost anywhere
between several hundreds of dollars to tens of thousands of dollars, with both X-Y plane pixel and Z-plane
layer resolutions as in the tens of microns, making them available on a larger scale not only for hobbyists,
but research groups such as ours as well. The DLP printer that the Malmstadt group purchased for this
project, from third-party vendor Proto Products, is the Asiga Max X27 UV.
1.1.1 Asiga Max X27 UV
The Asiga Max X27 UV was our primary apparatus for the fabrication of microuidic devices.
The Asiga Max X27 UV has a build volume of 113442 mm
3
, or approximately 7 in
3
; a volume a little bit
greater than that of four standard $10 bank rolls of quarters. This build volume is restricted though, to
the X-Y plane dimensions of 51.8 mm and 29.2 mm respectively, and a Z plane dimension restriction of
75 mm. The X-Y plane has a minimum square pixel resolution of 27m, a resolution that rivals even the
best SLA and DLP printers produced by research groups focused on improving SLA print resolution.
The Asiga Max X27 UV also has a Z-axis control sensitivity down to a 1m level, with a current Z-layer
minimum print height of 25m. The Asiga Max X27 UV’s projector operates at a ultraviolet wavelength
of 385 nm, and is a light emitting diode (LED) type light source.
The resin that we have consistently used is the complimentary transparent resin shipped with the DLP
printer, Pro3dure GR-1 Clear. The Pro3dure GR-1 Clear resin is known to be bio-compatible and its
biggest advantages are its transparency and structural strength after being fully cured. Though there are
specic resins including the Pro3dure GR-1 Clear which already have their information and parameters
input into the Asiga Max X27 UV already, the DLP printer also has a 100% Open Material System as well.
This indicates that this printer can print with 3
rd
party and self-developed resins as long the necessary
information and parameters of the resin are procured. Any device that has been printed has to be cleaned
with an appropriate solvent, in the case of Pro3dure GR-1 Clear it is isopropanol or isopropyl alcohol
2
Figure 1: Image of the Asiga DLP printer model Max X27 UV
(IPA), to fully remove any excess resin within the device or on its surface. This must be done before the
nal step curing the device because any resin left within or on the fabricated product will fuse to the
print during the curing process. This is important because any resin fusing within the inner channel of a
microuidic device will block and clog the device, causing it to malfunction.
1.2 Microuidics
Microuidics is the manipulation of a uid body in form factors with dimensions in the microm-
eter range or smaller. The study of microuidics can be dated back to as early as the 16
th
century, though
the term microfluidics was not rst mentioned and termed until 1989 by the Swedish pharmaceutical com-
pany Pharmacia Biosensor AB. Microuidics then became one of the hottest new technologies due to its
ability to produce methodologies that can manipulate low volume liquids, within the order range of 10
9
to 10
18
liters, in channels with dimensions on the order of 10
6
meters; a dimensional smaller than a
human hair[8]. The microuidic revolution did not reach its true potential though until after Defense
Advanced Research Projects Agency (DARPA) of the US government decided to invest in the technol-
3
ogy for the production integrated circuits. After its technological ascendancy, microuidics are now used
in a vast array of elds ranging from chemical analysis, food quality control, health diagnostics, to micro-
circuitry fabrication. Over the decades, the eld of microuidics has become one of the most technolog-
ically innovative elds of study that has produced devices such as lab-on-a-chip and even more recently,
organoids. Microuidic technologies has been aided by rapid developments in the production method-
ologies that have allowed for the rapid prototyping of microuidic devices. The original technology for
the production of microuidic devices was through photolithography, but that had many disadvantages
including being labor, time, and economically costly. With the development of poly(dimethylsiloxane)
(PDMS), scientists had an alternative fabrication method that is cheaper, less labor intensive, and has
great surface resolution, allowing for the explosion of research applying the new material. But PDMS
also has its own limitations; needing molds and clean rooms at times, prone to fracture depending on the
PDMS thickness and mixing ratio, and generally limited to 2-dimensional ow, new technologies such
as the previously mentioned SLA printing needs to be developed further. Some of the advantages and
disadvantages of microuidic technologies are also listed below,
Advantages Challenges
Small sample volumes (uL range) No set standards
Small devices total volumes Hard to scale up in the traditional sense
Micro-domain eects (Better Heat + Mass Transport) Lot of large external components needed
Customizable
Fast prototyping
T able 1: Advantages and challenges of microuidics
Some of the advantages of microuidics that we will be harnessing are the small sample volume needed,
as our samples are expensive, micro-domain eects where there is better homogeneity in heat and mass
transfer, and fast customizable prototyping. But like any other microuidic uidic system, we have our
share of challenges. Our devices are dependent on external ttings and large pumps which run our pro-
cesses.
4
1.2.1 Navier Stokes
Though analogs can be drawn in specic situations, he foundation of how microuidic systems
operate are fundamentally dierent from larger bulk systems. This dierences can be explained if we look
at a key principle of uid ow, the Navier-Stokes equation. T o qualify the Navier-Stokes equation, the
system must rst follow the continuity equation,

t
+O u = 0 (1)
If the continuity equation is qualied, then we can assume that we are dealing with an incompressible
uid, generally water or aqueous solutions with densities similar to water. Then we can write the full
Navier-Stokes equation as a system of time dependent vector elds, convective forces, pressure forces,
viscous forces, and gravitational body forces.

u
t
+ uOu

=OP +O
2
u +g (2)
For a steady state Navier-Stokes ow, all time dependent terms can be fully neglected as steady state means
that there is no change with respect to time. Gravitational terms can also generally be neglected as well,
due to dimensional analysis as the operating dimension is orders of magnitude larger than the direction
in which gravity operates. Lastly, convective terms, also known as inertial forces can be removed as vis-
cous forces dominate at the regime of low dimensions after applying dimensional analysis yet again. This
removes the nonlinear term, and the hardest term to solve for, from the Navier-Stokes equation. This
leaves us the resulting equation of,
0 =OP +O
2
u (3)
In our specic system, this system is actually analogous to what is a very well known system known as
pipe ow. Pipe ow, though not always, generally operates in cylindrical tubing, and therefore should be
used in cylindrical coordinates. P is the pressure in the system, with dimensions of Pascals (Pa). u is the
vector eld with dimensions of meters per second

m
s

. is the uid viscosity with dimensions of Pascal
5
seconds (Pa s). Within a pressure-driven system such as ours, this kind of pipe ow, or cylindrical ow
actually has a name, Hagen-Poiseuille ow. Hagen-Poiseuille ow is actually one of the few Navier-Stokes
systems that can be exactly solved with the no-slip condition at the wall of the tube, (r =R; u = 0), and
maximal ow at the center of the tube,

r = 0;
u
r
= 0

. This gives us the exact parabolic ow prole in
the radial direction,
u(r) =
R
2
4
P
x

1

r
R

2

(4)
This can further tell us that the maximum velocity at the center of the tube, wherer = 0,
u
max
=
R
2
4
P
x
(5)
T o further simplify, since Hagen-Poiseuille ow is a constant uidic system, with a constant pressure
dierential between the inlet and the outlet, the pressure term can be written as a pressure dierential over
a specic operating length and we can obtain exact pressure dierential in relation to the the volumetric
ow rate and the radius of the tubing.
P =
8QL
R
4
(6)
The pressure dierential, measured in Pascals, is the driving force in our system, and it is inversely pro-
portional to the radius of the tubing, measured in meters, to the fourth power. The pressure dierential
is also directly proportional to the volumetric ow rate, Q, which is measured in microliters per minute,

L
min

.
1.2.2 Reynolds Number
For a quick note as well, most if not all microuidic devices operate in the laminar ow regime. T o
determine if a ow is in the laminar regime, a Reynolds number must be calculated where in the inertial
forces must be compared to the viscous forces.
Re =
inertia
viscous
=
u
avg
L

(7)
6
The inertial forces is proportional to the density of the uid,, multiplied by the characteristic length
of the system, L, and the average velocity of the system, u
avg
. The viscous forces is only proportional to
the viscosity of the uid of the system,. As long as the Reynolds number in under 2100, we can say for
this specic system that the system is almost certainly operating in the laminar ow regime, where uid
vector lines do not cross each other. This also was another reason that we would have been able to remove
the nonlinear inertial term in Eq. 2. Our system is always operating in the laminar regime where viscous
forces dominate over inertial forces.
1.3 mRNA Display
mRNA display was rst developed in the 1990’s, and published in 1997, by the professor of our
collaborating group, Dr. Richard Roberts[9, 10, 11, 12]. mRNA display itself is a process that takes advan-
tage of a chemical linkage between a peptide and it’s encoding nucleic acid, a phenotype and a genotype
in a sense[13, 14, 15]. That is where in the genotype, in this case being the mRNA-cDNA complex, codes
for the the functional phenotype, in this case the attached peptide. If you have those two things linked,
you’re able to take pools with very high diversity, allow the peptides to bind to a specic desired target,
and whatever binds, you’re able to PCR and make a new fusion library with high anity for binding to
the target. There are six key steps to mRNA display as shown below in Fig. 2.
The rst step of mRNA display to to take a naive double-stranded DNA (dsDNA) library and transcribe
the library to get messenger RNA (mRNA) templates of each DNA strand. After transcription, the
unique step of mRNA display is the ligation step. During the second step, the ligation step, a puromycin
aminonucleoside attached to a DNA linker is ligated to the 3’ end of each mRNA strand. This can either
be done with or without a DNA splint to stabilize the complex. This is important, because this is what al-
lows us to create a peptide fusion during the translation step. During the third step, the translation step,
as the mRNA is being translated into a peptide, the corresponding polypeptides are covalently bound
to the puromycin. After which, during step four, the reverse transcription step, the complex fusion is
reverse transcribed to get the complementary DNA (cDNA) sequence and we are left with a library of
peptide-mRNA-cDNA fusions. The fusions are then useful because we can now perform the fth step
7
Figure 2: Schematic of mRNA display courtesy of Wan-Zhen Lin
of mRNA display with them: selection. Selection is the process by which fusions are directed towards a
target and selected based on their anity. During selection, as the fusions pass by a desired target which
has been immobilized onto a bead bed, dierent fusions will bind two the target based on their anities.
These fusion anities are based on their on-rates, or their binding coecients, but on-rates are diusion
limited processes. A fusion’s anity is determined by their o-rates, or their dissociation constants, as
those are not diusion limited. No matter how much or each fusion binds to the target, whether equimo-
lar or not, theoretically, the fusions or ligands with the best anity should have the lowest dissociation
constants relative to other fusions produced by the library. After selection is done, because the fusion
has mRNA complexed to cDNA, the fusion can undergo a polymerase chain reaction (PCR) to produce
dsDNA copies of the highest anity peptides from the original library. This creates a wholly new DNA
of solely high anity binders for the specic target we desire. But as you can see, this is a very complicated
process with each step having a large time delay, and selection taking the largest amount of time as gener-
ally many rounds of selection must be performed to get any sort of useful and narrow library. A typical
mRNA display cycle can take on average between 1-3 months, or even longer, to develop a useful library
of ligands[16]. That is why to help facilitate this process, we turn towards microuidics[17, 18].
8
2 Building the Automated System
Using microuidics, we set out to build an automated system for mRNA display with a syringe
pump that will use pressure as the driving force in the system. Here is an example attempt at incorporating
most of the mRNA display process steps into single automated system.
Figure 3: Example schematic of previously proposed automated mRNA display process courtesy of Wan-
Zhen Lin
Transcription, translation, and reverse transcription occur in an microuidic incubator above room tem-
perature in the device marked with a 1. Ligation occurs underneath a UV lamp at room temperature in
the device marked with a 2. Selection occurs in the devices marked with a 5. All of these dierent processes
are interconnected using 1/16-1/32” diameter polytetrauoroethylene (PTFE) tubing and polyether-ether-
ketone (PEEK) valves. A single round of mRNA display has the working liquid moving in this example
order:
• from the funnels into the incubator for transcription
• out from the incubator into the ligation device for ligation
• from the ligation device into the incubator for translation
• out of the incubator before going back into the incubator for transcription
• out of the incubator into the selection devices
These steps also don’t include heading back to the funnels to mix together with secondary chemicals
9
that cannot interact with previous chemicals until the working uid reaches that specic step. This also
disregards any purication steps in between to purify the working uid before moving onto the next step
so that a better yield may be obtained. An example of this would be after the ligation step, as ligation
has shown to have an empirical maximum yield of approximately 60%. The working uid then needs
to be puried before moving onto the next step, so that only the mRNA with puromycin attached to it
gets translated. If mRNA with puromycin attached to it is translated, the peptide-mRNA fusion would
not be created as there is no bridge to hold them together. All of this would not be possible without a
proper uid handling system that can accurately and precisely move the working uid from one part of
the system into another part of the system.
2.1 Fluid Handling
Initially the idea was that since we are using tubing, we can assume that ow through microuidic
tubing is very similar to another very well known and well studied phenomena: pipe ow. We proved
earlier in the background, Section 3.2.2 Navier-Stokes, that a ow equation and pressure dierential can
be solved exactly for this kind of a system. With an exact solution to such a ow equation, we should be
able to accurately determine where and how fast the uid is moving in the system. But alas, we CAN-
NOT use any of these equations because we break a very important assumption in Navier-Stokes, the
continuity equation. Besides having an incompressible liquid, such as water or other aqueous solutions,
in the system, we have a second HIGHLY COMPRESSIBLE ”liquid” in the system, air. Meaning all of
the assumptions that we have made are null and void. So what did we do? T o address the challenge of
uid handling and tracking then, we must turn towards placing sensors at key junctions of the system
and real-time tracking of the uid as it moves through the system.
10
Here is an example of sensor placements at the entrance and exit of some of the process devices.
Figure 4: Example placement of IR sensors of previously proposed automated mRNA display process
courtesy of W an-Zhen Lin
The sensors placement would be able to inform us on whether the working uid had directly entered or
exited the device. The type of sensors that we opted to use for this system were infrared sensors with a
transmitter and receiver as depicted.
Figure 5: Example schematic of a IR sensor apparatus courtesy of Wan-Zhen Lin
The transmitter and receiver never have any direct contact with the working uid. A small device can be
fabricated via stereolithography such that the middle is a narrow channel that the tubing can slot right
through. Then two cylindrical orices are also fabricated perpendicular to that of the tubing so that the
transmitter and receiver have a direct line of sight to each other. T o understand further how these infrared
11
sensors work, please see the diagram below in Figure 6. As you can see in the rst tubing diagram below,
with uid moving in the x direction, at time zero before the uid has not reached the sensor, the signal
that the receiver reads from the transmitter is stable because it is unobstructed.
Figure 6: Example schematic of how IR sensors operate
Some time later in the second tubing diagram, as uid moves in front of the receiver, the signal decreases
because the signal is refracted and absorbed by the working uid and thereby not reaching the receiver.
Finally as more time has passed in the third tubing diagram, and the working uid fully moves past the
receiver, the signal increases as the signal from the transmitter is no longer obstructed. We used this spe-
cic behavior and that fact that the IR sensors are constantly outputting a signal to write a Python code
for the real time monitoring and foundational uid handling of the reactants and working uids in the
automated mRNA display system. This process only works if the owrate is small and there are no build
ups or constrictions within the system. As mentioned previously, this is a pressure driven system and
there is a compressible uid in the system. At high ow rates, the rapidly changing volume of the com-
pressible uid can drastically aect the speed and behaviour of the incompressible working uid. But as
long as the ow rate is suciently slow enough and dimensionality is conserved within the system, the
IR sensors can handle the compressible uid’s aects on the system. The most important and necessary
place to start the project was to make sure that the working uids gets to where it needs to go.
12
3 Microuidic O-Rate Selections
After we have the foundations for being able to move liquids throughout the system accurately and
precisely, we can start to look at each step of the process in isolation to see how we would turn the manual
process into a microuidic process. One of the steps that we had the most success in the microuidic
conversion, was actually was actually the selection step. For the Selection step, we were able to develop a
microuidic platform for the enrichment of higher anity ligands over lower anity ligands.
3.1 MFED
We call this device the MFED, or the Microuidic enrichment device. The MFED consists of a
porous frit that is held in place by PTFE tubing, a microuidic casing, and a bead bed loaded onto one
side. This can be seen in the schematic of Figure 7 below.
Figure 7: Schematic of how an MFED operates
The MFED functions such that a continuous ow is used to load a sample, which binds to the desired
target that has been immobilized on the bead bed. The MFED can also be used to facilitate continuous
washing of the bead bed, to carry away any undesired low anity peptides that have dissociated away
from the target. Based on that description, the MFED is almost like a smaller, cheaper homemade cousin
to a chromatography column, except where as in the case of a chromatography column, only the eluent
is generally desired, for the MFED, the eluent can be desired after a purication step AND the beads
are desired because the desired ligands are bound to the immobilized target. The MFED operates using
13
similar principles to that of a chromatography column where residence time of the uid in the bead bed
plays a large role in the binding mechanisms of the device. This idea of residence time and its eects on
the binding mechanisms of the system will be explored further in a series of experiments.
Seen below are images of the nal prototype MFED after more than 4 iterations.
Figure 8: Picture of an MFED deconstructed
Figure 9: Picture of an MFED with blue liquid owing through it as a reference
As you can see by the images, there is an inner recess in the microuidic housing for the frit to t snugly
into. Any pore sized frit can be stabilized in the microuidic housing as long as it has the same outer
diameter as the 10m frit that is pictured. If the outer diameter of the frit diers in any way, that is actually
an easily solvable problem due to the nature of rapid stereolithographic prototyping. A new housing
design is easy to make, and the fabrication of the device is relatively quick as well due to the inherently
small size of the microuidic device. The pictured MFED devices each take less than 3 hours to print using
our Asiga printer, and multiple can be printed at once, allowing for the ease of parallel testing of the new
device. The true bottleneck comes in the connectors for the tubing. Due to the unstandardized nature
14
of microuidic device fabrication, external components are deep in both breadth and depth. This allows
for commercial microuidic companies to charge exorbitant prices for the pass-through connectors and
ferrules that are pictured with the MFED. B
3.2 Selection Experiments
T o test the MFED, we did a series of selection studies. We used the marker BcL-xL as the immo-
bilized target and two previously identied ligands: E1 a higher anity ligand, and Pep2 a lower anity
ligand, as you can see by their dissociation constants in the table below, which are nearly 4 orders of mag-
nitude apart.
E1 7.4 x 10
6
s
1
Pep2 1.1 x 10
2
s
1
T able 2: T able of dissociation rate constants k
o
.
For the rst selection study we did, we needed to nd suitable ow rate for sample binding, and washing.
T o do so, we set up an experiment to measure the percentage of ligand that binds to the immobilized target
on the beads with respect to a constant ow rate. These owrate vs. binding percentage experiments were
done solely using the higher anity ligand, E1 in a single pass-through continuous ow for binding. Our
results are graphed on the gure below along with a model that has been tted to the data. From this data
and model, we can see that the binding mechanism of the ligand to the bead bed is similar to the E1-Bclx
L
formation reaction. As the ow rate decreases and therefore the residence time in the bead bed goes up,
there is an increase in the binding.
This is also inversely true where higher ow rates mean lower residence times in the bead bed, correlat-
ing to lower binding. The way that residence time was calculated is slightly dierent than normal ow
reaction systems.
=
V (1
Q
) (8)
15
Figure 10: Characterizing and modeling E1 peptidemRNA fusion binding at various ow rates. The
percent bound increases as the owrate decreases and plateaus at 73%. Error bars represent the standard
deviation of percent bound over three trials. The best-t model gives a k
on
= (2:5 4)x10
4
M
1
s
1
Normal residence time calculations would just have volume divided by the average volumetric ow rate
of the system, but we have a 1 modier because we are owing through a porous media, and need
to take that into account. Flowing through a porous media means that only a fraction of the bead bed
is vacant for uid to move through it, resulting in being the packing eciency of the bead bed, with
perfect spherical bead packing being 75%. In our model, we were able to t a pseudo rst-order on-rate
model to the data to determine a packing eciency of 74%, a resident time of 6 seconds, and an on-rate
of (2.5 0.4) x 10
4
M
1
s
1
, which agrees well with published data on the diusion limit of ligands. The
pseudo-rst-order on-rate model we used was
B =B
max
(1e
k on[T]
0

) (9)
where B is the ligand percentage bound, and the two known values of the system being B
max
is the maxi-
mum amount of ligand experimentally found to have bound, and [T]
0
is the initial target concentration.
Though B
max
was nowhere near 100%, this can be explained by errors in the mRNA template construc-
16
tion. With 63 positions for out specic sequence, a 75% error corresponds to about a 0.5% errate rate per
position: (0.995)
63
= (0.73). Using this knowledge selected 25 uL/min as the binding ow rate because
as it is the inection point in the graph, any slower and it would not be justied in spending more time
to bind the sample for the moderate increase we get in binding percentage. That would give us a res-
idence time of approximate 3 seconds in the bead bed and binding percentage of nearly 60%. We also
selected 500 uL/min as the washing ow rate as that is where the graph is starting to plateau, and any
faster and we may start running into pressure issues and wasting washing buer. That would give us a
binding percentage of nearly zero because the residence time of the working uid in the bead bed is only
150 milliseconds.After using the rst set of experiments to select the binding and washing owrates, we
conducted a set of washing experiments to predict the enrichment ratio between E1 and Pep2. In these
experiments, each ligand was tested in separate experiments, where they were bound to beads at the bind-
ing ow rate, then washed for a specic set of time intervals. Afterwards, the data is plotted as you can
see, with the ratio of E1 to Pep2 calculated from the percentage of each individual ligand left on the beads
at each specic wash interval.
Because the E1 binding data is relatively uninteresting due to it’s plateauing nature at a high binding
percentage, we focused on modeling the binding for Pep2. Loading at 25 uL/min, Pep2 was able to bind
approximately 12% of itself to the Bcl-xL. Using binding percentage again, we modeled the dissociation
of Pep2 from the target using a rst order model.
B = (B
0
B
min
)e
k
o
w
+B
min
(10)
where the known variables are B
0
, the initial amount bound, B
min
, the minimum amount of target that
must be bound, and
w
is the wash time, a variable which we set. We can then model the data and deter-
mine a what the experimental k
o
is. The k
o
was found to be 9 x 10
3
s
1
, which agrees very well with the
manual result found, 1.1 x 10
2
s
1
.Ideally using this data, we would select a wash time that maximizes the
enrichment of higher anity ligands over lower anity ligands, in this case E1 over Pep2. As we can see,
a wash time of approximately 15 mins is sucient, because since the trend is always going upwards, even
though we want to maximize enrichment, which would be at the far right end of the graph, the trade o
17
Figure 11: MFED loading and washing for E1 (magenta) and Pep2 (green) mRNApeptide fusions. Error
bars indicate the standard deviation of three trials. Initial binding for E1 (52%) shows little decay during
washing, whereas Pep2 binding (12 1%) washes out with a rst order rate constant k
o
= 92x10
3
s
1
.
The ratio of E1 to Pep2 bound (blue) plateaus after approximately 1000 seconds when the Pep 2binding
reaches background binding (1.4 0.3%)[19]
for the amount of time spent washing for the minimal increase in enrichment ratio is no longer worth it
as it starts to plateau. In the set of experiments we just talked about, we were predicting the enrichment
ratios of higher anity ligands over lower anity ligands.
18
In this nal set of experiments we did with MFED, we actually mixed dierent molar ratios of E1 and Pep2
and enriched them both manually and with the device. As seen on the chart below, using the MFED we
were able to statistically enrich E1 over Pep2 better than with manual selection at every molar ratio, even
the most dilute sample with a molar ratio of 1:292.
Figure 12: Manual versus MFED enrichment of E1 over Pep2. Selection was performed on various starting
ratios of E1/Pep2 (1:2,1:38, 1:67, 1:135, and 1:292). MFED selection consistently outperformed manual
selection, averaging 13-fold enrichment compared to 5-fold enrichment observed with manual selection
(*indicates p< 0.05)[19]
19
You can also visually see this in the gel image above after PCR at both the 1:38 and 1:67 molar ratio where
the MFED bands are similar in brightness, where as the manual selection bands has Pep2 brighter than
E1 in both.
Figure 13: Manual versus MFED enrichment of E1 over Pep2. Selection was performed on various starting
ratios of E1/Pep2 (1:2,1:38, 1:67, 1:135, and 1:292). Resulting PCR products were run on agarose gels. Gel
images were analyzed to extract molar ratios before and after selection.[19]
The MFED has shown to be a device that works better at enriching higher anity ligands than the manual
technique. The results of these MFED experiments were published recently in the Journal of Analytical
Chemistry[19].
20
4 Challenges
But where we have success, we will also always have challenges. When we were looking to to con-
vert the rst step of the mRNA display process, transcription, using microuidics, we ran into some
diculties.
4.1 Protein Sticking
Figure 14: CAD drawing of the incubator
Above is a modeled 3D CAD drawing of one of the many incubator prototypes of that were able
to rapidly print with our Asiga printer. As you can see in the gel image below, during transcription the
yield from the device was consistently lower than that of the manual technique with faint bands for the
device trials and dark bands for the control trials. We pondered why this was happening and conducted
a series of experiments for transcription, controlling for one variable at a time.
Figure 15: This is seven transcription tests that have been run on a urea gel. The ve tests on the left were
incubated in the microuidic incubator. The two tests were incubated in eppendorf tubes on a heating
block.
21
4.1.1 Transcription Tests
After theorizing that it may be the device interacting with the transcription protein T7, we con-
ducted a series of experiments as seen below where each solution had dierent levels of contact with the
incubator. The rst sample had no contact with the incubator, with all the reagents premixed in an ep-
pendorf tube before being place on a standard heating block. That sample performed best. The second
sample had minimal contact with the microuidic incubator, with all the reagents premixed in an eppen-
dorf tube before being incubated in the microuidic incubator. The second sample’s yield suered. The
third sample only had contact with microuidic incubator, being mixed in it, and also being incubated
in it. The yield for the third sample was basically nonexistent.
Figure 16: Design of experiment was used to determine the root cause of the low transcription yield. The
left trial had the zero contact with the microuidic incubator. The middle trial had minimal contact with
the microuidic incubator. The right trial only had contact with the microuidic incubator.
T o conrm that it was the T7 interacting with the microuidic incubator, for the experiment below, in
the third sample, we owed all the reagents for the reaction, the DNA and transcription buer, sans the
T7, in and out of the incubator and then spiked in the T7 before incubating it on the standard heating
block. The yield compares favorably to that of the manual control sample which is the left sample, and
the middle sample is a sample which was mixed and incubated in the microuidic device.
22
Figure 17: Gel analysis to conrm transcription protein deactivation. The left trial was a control trial.
The middle trial only had contact with the microuidic incubator. The right trial had all the components
touch the incubator besides the transcription protein touch the incubator.
4.1.2 Fluorescent Microscopy
T o double check, we conducted a set of uorescent microscopy tests to conrm our suspicions.
NHS-rhodamine was the uorescent dye we used to determine the positioning of protein within the
system. The NHS-rhodamine was ligated using DMF and the protocol provided by the manufacturer,
ThermoFisher. The dye was passed through a lter after ligation to get rid of any excess dye in solution
as that would aect uorescent measurements. Below, we have the negative control tests where only the
uorescent dye NHS-rhodamine was initially owed through the system. As you can see in uorescence
plot below and somewhat visually in the channel, there is a VERY small signal, with 3 distinctive bumps.
After 5 washes with PBS (phosphate buered solution), the uorescence plot becomes almost entirely
at, and visually, the channels look like the background signal as it is indistinguishably blurry through
the microscope lens now.
23
Figure 18: Right, negative control where only uorescent dye, NHS-rhodamine, was owed through the
channel. Left, negative control after 5 washes of PBS
For the positive control test, NHS-rhodamine is ligated to the protein T7 and then owed thorough the
device. On the third image, you can see signicantly more uorescence than with just the dye alone.
There are almost crystal like patterns of uorescence. But that likely just tells us that the inner channel is
quite rough, and the protein is sticking to all surface, even the crevices. After 5 washes of PBS not only
did uorescence persist, besides one of the channels, the signal strength and shape basically didn’t change.
Washing had almost zero eect on the sticking of the proteins. This is conclusive evidence that the T7
protein is interacting with the walls of the device and getting stuck there.
But now the question is how do we solve the problem?
24
Figure 19: Right, positive sample with transcription protein T7 ligated with uorescent dye, NHS-
rhodamine, was owed through the channel. Left, positive sample after 5 washes of PBS
]
4.2 Resin Development
T o address this challenge in protein sticking, we decided that a change in the resin material may be
needed. Therefore, we turned towards other groups that have been working to develop bio-compatible
resins. One of the most common polymers that has been known to be bio-compatible is polyethylene
glycol diacrylate, or PEG-DA.
Figure 20: Molecular schematic of PEG-DA
25
This polymer along with PDMS are some of the most used materials in developing biocompatible sys-
tems, except people have not been able to turn PDMS into a printable resin, but that is not the case for
PEGDA.
4.2.1 Recipe Selection
We based our resin on a paper that came out by Kuo et al., out of Korea in 2019, trying to replicate
their formula using Bis(2,4,6-trimethylbenzoyl)-phenylphosphineoxide, or irgacure-819 [20],
Figure 21: Molecular schematic of IRG-819
, and 2-isopropylthioxanthone or ITX.
Figure 22: Molecular schematic of ITX
We decided to follow their formula of using IRG-819 as the photoinitiator and ITX as the UV absorber,
because the absorbance spectrum of these two chemicals lands neatly in the 385 nm range, exactly that of
the Asiga printer that we have. Luckily our Asiga accepts third party materials, so after mixing the resin,
the next thing we did was test it.
4.2.2 Resin Testing
T o print with the PEGDA resin, we rst needed to get some parameters about the resin. We needed
to nd out relationship between how much energy is used to print, and the resulting layer height. T o nd
out, we created an apparatus that consists of a glass slide bonded to a chunk of PDMS with a well cut out
of it for where the resin would sit. The bottom of the glass slide was covered such that UV light would
26
only shine unidirectionally upwards through the well, and not refract in from the sides and create solids
from the side. W e chose to create the standard of measurement with respect to a set interval of energy
input into the resin.
Figure 23: PDMS well on top of a glass slide
The resultant resin layers that were produced were then measured by hand caliper if thick enough, but
for the thinner slices, a prolometer needed to be used due to their margin of error and frailty as the layer
heights measure in the nanometer range at this point. Below is how a prolometer works, very similarly
to an atomic force microscope, or AFM.
Figure 24: Schematic of how a prolometer operates
The data was plotted on the right against the data from clear commercial Asiga resin Pro3dure GR-1 Clear
and then input into the printer so it had to the necessary information needed to print. As you can see, the
new PEG-DA resin needs more energy than the Pro3dure GR-1 Clear resin to print larger layers, while
27
also plateauing much earlier as well. This may be due to the yellow pigment of the resin due to the yellow
nature of IRG-819. This causes the diraction of the light energy and cannot be absorbed properly. We
also have a very powerful UV absorber in the resin, ITX.
Figure 25: Layer height vs. Light Source Energy graph for the Pro3dure GR-1 Clear and homemade PEG-
DA resins
Now that the Asiga printer has the necessary parameters needed to print using the new resin, we were able
to print some small posts using the the PEGDA. We also printed similar posts with the clear commercial
Asiga resin Pro3dure GR-1 Clear for comparison. We then submerged the printed parts in separate ep-
pendorf tubes full of dilute T7 taken from a control stock. After being incubated for an hour we ran the
solutions using a Pierce BCA assay to quantify the amount of protein left in solution.
Initial Pro3dure GR-1 Clear PEG-DA
Protein Conc. (nM) 461.6 310.9 348.6
% lost 32.6% 24.5%
T able 3: Resin comparison
The BCA assay showed that the stock solution initially had a protein concentration of approximately
462 nM. The amount of protein left in the Asiga Resin Pro3dure GR-1 Clear and PEGDA were approxi-
mately 311 nM and 349 nM respectively. That means that aproximately 33% of the proteins stuck to Asiga
Pro3dure GR-1 Clear, while only 25% of the proteins stuck to PEGDA. You can also phrase that as say
that the PEGDA outperformed the Asiga Pro3dure GR-1 Clear by 8% overall, or 25% relatively.
28
5 Current Projects
So, we were able to moderately improve the material that consists of all our devices in the mRNA
display system. That helps our entire mRNA display system in general as our overall goal of producing
a complete run of mRNA display from transcription through selection has still not been met, and we
are still working towards that. Each of the individual steps of mRNA display have been reproduced
microuidically in isolation, and the automated uid handling has shown to be able to complete the
entire process using water as a stand-in. Now our current project is to make it all work in tandem.
Figure 26: A SEM image of non-porous silica beads.[21]
Another interesting current project of note, is that when we were doing the MFED washing trials, for the
low anity binders, the ligand percentage bound did not trend towards zero as expected. It plateaued
right above zero, at approximately 1.3%, which is not negligible. We assume that ligands are likely being
trapped within the bead matrix as we are using porous agarose beads similar to what you can see in the
bottom right. W e propose and hypothesize that if you used non-porous beads like silica with target bound
to them, we should not see that same problem. Though, with non-porous beads, we may get lower total
binding as there is less surface area and therefore less reactions sites, that should be ne as long as the low
anity ligands don’t bind. Those trials are still yet to be nished, though we have the necessary materials
already.
29
6 Conclusions
In conclusion, while trying to develop an automated system for mRNA display, we had both suc-
cesses and challenges. W e able to develop a new method and platform for doing selection that works
better than the manual method. W e even got a paper out of it, where I was second author behind the
two co-rst authors. W e were able to correctly theorize and prove that the T7 proteins were interacting
and sticking to the chamber walls. W e addressed the problem with a moderate degree of success. And we
are step by step getting closer and closer to to our ultimate goal of having an automated mRNA display
system.
30
7 Future Directions
Some future directions that this project may take though is that while we did have moderate success
in nd a resin that proteins stick less to, what would be best is if we can develop a bio-inert resin rather than
a bio-compatible resin. That way we have no interaction with the walls whatsoever. A good candidate
for that is to use a uoropolymer like the one used to make PTFE tubing. Another direction the project
can take is that while loading the non-porous beads as a test, because the silica beads have such a small
diameter, a much smaller pored frit needed to be used in the MFED. This led to a large pressure drop in
the tubing due to restricted ow, and the velocity prole was visibly sharpened where the length of the
prole was tens of times greater than the witdth of the prole. This may have implications in sheath ows
that may not need a secondary immiscible carrier uid to focus the ow to a precise central location.
31
References
[1] A. Venkataraman, K. Y ang, J. Irizarry, M. Mackiewicz, P. Mita, Z. Kuang, L. Xue, D. Ghosh, S. Liu,
and P. Ramos. A toolbox of immunoprecipitation-grade monoclonal antibodies to human tran-
scription factors. Nat. Methods, 15:330, 2018.
[2] R. Doshi, B. R. Chen, C. R. T. Vibat, N. Huang, C. W. Lee, and G. Chang. In vitro nanobody
discovery for integral membrane protein targets. Sci. Rep., 4:6760, 2015.
[3] E. T. Boder, K. S. Midelfort, and K. D. Wittrup. Directed evolution of antibody fragments with
monovalent femtomolar antigen-binding anity. Proc. Natl. Acad. Sci. U. S. A., 97(20):10701, 2000.
[4] J. Qian, X. Lou, Y. Zhang, Y. Xiao, and H. T. Soh. Generation of highly specic aptamers via
micromagnetic selection. Anal. Chem., 81(13):5490, 2009.
[5] S. S. Oh, K. M. Ahmad, M. Cho, S. Kim, Y. Xiao, and H. T. Soh. Improving aptamer selection e-
ciency through volume dilution, magnetic concentration, and continuous washing in microuidic
channels. Anal. Chem., 83(17):6883, 2011.
[6] X. Lou, J. Qian, Y. Xiao, L. Viel, A. E. Gerdon, E. T. Lagally, P. Atzberger, T. M. T arasow, A. J.
Heeger, and H. T. Soh. Micromagnetic selection of aptamers in microuidic channels. Proc. Natl.
Acad. Sci. U. S. A., 106(9):2989, 2009.
[7] K. C. Bhargava, R. Ermagan, B. Thompson, A. Friedman, and N. Malmstadt. Modular, discrete
micromixer elements fabricated by 3d printing. Micromachines, 8(5):137, 2017.
[8] George M. Whitesides. The origins and the future of microuidics. Nature, 442(7101):368–373,
2006.
[9] R. Liu, J. E. Barrick, J. W . Szostak, and R. W. Roberts. [19] optimized synthesis of rna-protein
fusions for in vitro protein selection. Methods Enzymol., 318:268, 2000.
[10] F. Jalali-Y azdi, T. T. T akahashi, and R. W. Roberts. General, label-free method for determining k d
and ligand concentration simultaneously. Anal. Chem., 87(23):11755, 2015.
[11] F. Jalali-Y azdi, L. Huong Lai, T. T. T akahashi, and R. W. Roberts. High-throughput measure-
ment of binding kinetics by mrna display and next-generation sequencing. Angew. Chem., Int. Ed.,
55(12):4007, 2016.
[12] F. Jalali-Y azdi, J. M. Corbin, T. T. T akahashi, and R. W. Roberts. Robust, quantitative analysis of
proteins using peptide immunoreagents, in vitro translation, and an ultrasensitive acoustic resonant
sensor. Anal. Chem., 86(10):4715, 2014.
[13] H. I. Liao, C. A. Olson, S. Hwang, H. Deng, E. Wong, R. S. Baric, R. W. Roberts, and R. Sun.
mrna display design of bronectin-based intrabodies that detect and inhibit severe acute respiratory
syndrome coronavirus nucleocapsid protein. J. Biol. Chem., 284(26):17512, 2009.
[14] E. T. Boder and K. D. Wittrup. Optimal screening of surface-displayed polypeptide libraries. Biotech-
nology progress, 14(1):55, 1998.
32
[15] M. Cetin, W . E. Evenson, G. G. Gross, F. Jalali-Yazdi, D. Krieger, D. Arnold, T. T. T akahashi, and
R. W. Roberts. Rasins: genetically encoded intrabodies of activated ras proteins. J. Mol. Biol.,
429(4):562, 2017.
[16] L. Xiao, K. C. Hung, T. T. T akahashi, K. I. Joo, M. Lim, R. W. Roberts, and P. Wang. Antibody-
mimetic ligand selected by mrna display targets dc-sign for dendritic cell-directed antigen delivery.
ACS Chem. Biol., 8(5):967, 2013.
[17] Y. Liu, J. D. Adams, K. Turner, F. V. Cochran, S. S. Gambhir, and H. T. Soh. Controlling the
selection stringency of phage display using a microuidic device. Lab Chip, 9(8):1033, 2009.
[18] C. A. Olson, J. D. Adams, T. T. T akahashi, H. Qi, S. M. Howell, T. T. Wu, R. W. Roberts, R. Sun,
and H. T. Soh. Rapid mrna-display selection of an il-6 inhibitor using continuous-ow magnetic
separation. Angew. Chem., Int. Ed., 50(36):8295, 2011.
[19] William E. Evenson, W an-Zhen Sophie Lin, Kenmond Pang, Alexander T. Czaja, Farzad Jalali-Yazdi,
T erry T. T akahashi, Noah Malmstadt, and Richard W. Roberts. Enabling ow-based kinetic o-
rate selections using a microuidic enrichment device. Analytical Chemistry, 92(15):10218–10222,
08 2020.
[20] Alexandra P. Kuo, Nirveek Bhattacharjee, Yuan-Sheng Lee, Kurt Castro, Yong T ae Kim, and Albert
Folch. High-precision stereolithography of biomicrouidic devices. Advanced Materials Technolo-
gies, 4(6):1800395, 2019.
[21] Jihee Kim, Prakash Chandra, Jiyoon Yang, and Seog Woo Rhee. Growth and migration of balb/3t3
broblast cells on nano-engineered silica beads surface.
33

Abstract (if available)

Abstract Potential targets for new therapeutics and diagnostic tools are being discovered at a greater rate than existing methods are able to determine ligands that can bind to these targets[1, 2]. New methodologies and devices need to be developed to be able to match the pace in which targets are discovered and one of those pathways is via stereolithographic printing[3, 4, 5, 6]. We have used stereolithographic 3-D resin printing to rapidly prototype bio-compatible microfluidic devices to facilitate an automated system for mRNA display. Automating and developing microfluidic analogs of mRNA display process will drastically reduce the sample size, process cost, and total time needed to process a target[7]. One of the most successful devices we developed and stereolithographically produced, is the microfluidic enrichment device (MFED) which enables kinetic off-rate selection of ligands without needing an exogenous competitor. Simple and known kinetic equations are also able to model ligand binding and dissociation in the MFED, with experimental rate constants agreeing with other independently measured constants. After MFED variables were tuned such that ligands bind and dissociate to a selected target based on affinity, we demonstrated that the MFED can enrich a ligand at least 4 times greater than that of a standard manual selection. While developing the MFED, there were challenges in the production of bio-compatible microfluidic devices due to protein sticking (adsorption) to device surfaces, specifically the inner channel walls. This was discovered by using simple design of experiment and confirmed by fluorescent microscopy. To address this problem, new bio-compatible resins were developed and experimented with, to determine their efficacy and suitability for the mRNA display automation project. The overall goal of the project to automate mRNA display is a lofty goal that is still unaccomplished, but with interesting challenges and side projects solved, we are step by step attempting to complete that goal.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Intricate microfluidic devices for biopharmaceutical processes: forging ahead with additive manufacturing

PDF

The selective targeting of Ras: a tale of kinetics, mRNA display, and vaccines

PDF

Synthesis of high-quality nanoparticles using microfluidic platforms

PDF

Generation and characterization of peptide theranostics by mRNA display

PDF

The patterning of polymer thin films on porous substrates via initiated chemical vapor deposition

PDF

Predictable microfluidic mixing using discrete element microfluidics

PDF

Vapor phase deposition of dense and porous polymer coatings and membranes for increased sustainability and practical applications

PDF

Scale-up of vapor-phase deposition of polymers: towards large-scale processing

PDF

Fabrication of functional porous membranes via polymerization of solid monomer by a vapor-phase initiator

PDF

Fabrication of polymer films on liquid substrates via initiated chemical vapor deposition: controlling morphology and composition

PDF

Expanded functionality and scalability of modular fluidic and instrumentation components

PDF

Customization of the mRNA display cycle – from protease resistance to high throughput analysis of peptide ligands

PDF

Chemical recycling of amine/epoxy composites at atmospheric pressure

PDF

High-throughput nanoparticle fabrication and nano-biomembrane interactions

PDF

Continuous flow synthesis of catalysts with custom made reactor with flow and temperature studies

PDF

Engineering therapeutics for the improved antitumor efficacy of chimeric antigen receptor T cell therapy

PDF

Formation of polymer gels, films, and particles via initiated chemical vapor deposition onto liquid substrates

PDF

A modular microscale laboratory

PDF

Developing a method for measuring kinetic rates of RNA/protein interaction using switchSENSE® technology on the DRX2

PDF

Development of fabrication technologies for robust Parylene medical implants

Asset Metadata

Creator Pang, Kenmond Hall (author)

Core Title Rapid microfluidic prototyping for biological applications via stereolithographic 3-D printing: a journey

School Viterbi School of Engineering

Degree Master of Science

Degree Program Chemical Engineering

Publication Date 12/13/2020

Defense Date 10/29/2020

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag 3-D printing,automation,microfluidics,mRNA display,OAI-PMH Harvest,resin,stereolithography

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Malmstadt, Noah (committee chair), Gupta, Malancha (committee member), Roberts, Richard (committee member)

Creator Email kenmond.pang@gmail.com,kenmondp@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-408358

Unique identifier UC11667412

Identifier etd-PangKenmon-9217.pdf (filename),usctheses-c89-408358 (legacy record id)

Legacy Identifier etd-PangKenmon-9217.pdf

Dmrecord 408358

Document Type Thesis

Rights Pang, Kenmond Hall

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA