Controlling the selection stringency of phage display using a microfluidic
Yanli Liu,abJonathan D. Adams,cKelisha Turner,eFrank V. Cochran,fSanjiv Sam Gambhirfand
H. Tom Soh*bd
Received 24th November 2008, Accepted 18th February 2009
First published as an Advance Article on the web 3rd March 2009
and demonstrate that accurate control of washing stringency in our
of isolated peptide sequences. Reproducible generation of magnetic
and fluidic forces allows controlled washing conditions that enable
rapid convergence of selected peptide sequences. These findings may
provide a foundation for the development ofautomated microsystems
for rapid in vitro directed evolution of affinity reagents.
The invention of phage display technology1established the paradigm
an indispensable tool for probing specific biochemical molecular
interactions.2–4In this method, a library of peptides or proteins,
typically consisting of ?109members, is expressed on the outer
surface of a bacteriophage with the DNA encoding each variant
contained within the virus particle.5This physical linkage between
each peptide sequence and its encoding DNA allows rapid in vitro
selection of molecules that bind to a target of interest with high
affinity and specificity.
Typically, the selection is performed through a ‘‘panning’’ process
inwhichthetargetmoleculesareimmobilizedon asolid support(e.g.
Petri dishes, microtiter plates or magnetic beads), incubated with the
The outcome of the phage selection critically depends on the strin-
gency conditions imposed during the panning process. There are
many factors that govern the stringency including concentrations of
the target molecule and the phage library, incubation time, temper-
ature, pH, salt concentration and washing conditions.4,5,9Conven-
tional methods of panning have proven fairly effective, but suffer
from a few disadvantages. First, they require significant amounts of
when the target is not abundantly available.5Second, the process
often yields phage binders that interact with the solid support rather
than the target, necessitating negative selection steps.10Finally, it has
proven challenging to reproducibly control the washing conditions,
which has a direct impact on the resulting clones that are isolated.9,11
Microfluidic technology offers many unique advantages for
molecular separations.12–17Here, we present the first application of
microfluidics for phage selection. Using streptavidin (SA) as a model
target molecule, we demonstrate micro-scale control of washing
stringency, and show that this has a direct impact on the efficiency of
identifying peptide binding motifs. Finally, we show that such
microfluidics-based systems can enable the development of ultra-
rapid, highly efficient and automated platforms for the generation of
affinity reagents in a miniaturized format.
We adopted the Micro-Magnetic Separation (MMS) device
(Fig. 1(a)) for the panning process. The detailed fabrication
procedure of the device has been previously described by our
separated by a 30 mm thick polymer layer, with a set of patterned
nickel structures on the bottom glass layer. The microfabricated
releaseof magneticbeads,which isessential forcontrolled washing
between nickel and the buffer medium (mr,nickel¼ 200, mr,buffer?1),
the placement of a neodymium–iron–boron (NeFeB) external
magnet (grade N42, K&J Magnetics, Jamison, PA) on the chip
results in the automatic and reproducible generation of large
at the edges of the nickel patterns (Fig. 1(b), left).20The magnitude
of the gradient is approximately 104T/m within 8 mm of the struc-
tures and the resulting magnetophoretic force (~ Fm) is tens of
nanonewtons as approximated by~ Fm¼ mVB, where m is the
magnetization of the bead.21During the washing step, we ensure
that the fluidic drag force (~ Fd) is less than~ Fmso that the beads do
typically less than 10 pN under our experimental conditions as
approximatedby~ Fd¼ 6pha~ vfwherehisthefluidviscosity,aisthe
removal of theexternal magnet de-magnetizes thenickel structures
so that the beads are effectively eluted from the chip (Fig. 1(b),
right). Under experimental conditions, we found that the magnetic
beads were indeed tightly bound to the magnetic traps during the
washing steps, and we were able to recover 99.5% of the beads that
entered the device as measured via flow cytometry.
We performed our phage selection with Dynabeads? M-270
magnetic beads coated with streptavidin (SA), purchased from
Invitrogen (Carlsbad, CA). The PhD-7 phage display peptide
library containing ?109unique sequences was purchased from
New England Biolabs (Ipswich, MA). We washed the magnetic
beads twice for 5 min with tris-buffered saline (TBS, 50 mM Tris-
HCl,150 mMNaCl,pH7.5),then resuspended10 mlof thewashed
aNeuroscience Research Institute, University of California, Santa Barbara,
CA 93106, USA
bDepartment of Materials, University of California, Santa Barbara, CA
93106, USA. E-mail: firstname.lastname@example.org; Fax: +1-805-893-8651;
cDepartment of Physics, University of California, Santa Barbara, CA
dDepartment of Mechanical Engineering, University of California, Santa
Barbara, CA 93106, USA
eSchool of Science and Technology, Jackson State University, Jackson,
MS, 39217, USA
fMolecular ImagingProgram at Stanford, Departmentof Radiology& Bio-
X Program, Stanford University, Stanford, CA, 94305, USA
This journal is ª The Royal Society of Chemistry 2009 Lab Chip, 2009, 9, 1033–1036 | 1033
COMMUNICATIONwww.rsc.org/loc | Lab on a Chip
beads(ataconcentrationof? 7? 108beads/ml)in80mlofTBS,to
which 10 ml of the library containing 2 ? 1011phage was then
with gentle agitation (Fig. 1(c), step A). Next, NeFeB permanent
magnets were placed on the MMS device to establish the magnetic
field gradients inside the microchannel (Fig. 1(c), step B). Two
programmable syringe pumps (PhD 2000, Harvard Apparatus,
Holliston, MA) delivered sample and buffer into the device at flow
beads were washed under a variety of conditions. Finally, the
external magnets were removed from the device to de-magnetize
the nickel patterns, and the bead-bound phage were eluted at 50
ml/h. The device was then sequentially washed with 2 ml of 25%
bleach, 5 ml of DI water and 5 ml of TBS buffer for later reuse.
We eluted the bound phage from the magnetic beads by incu-
bation with 0.1 mM biotin in TBS for 30 min within a magnetic
particle concentrator (Dynal MPC-15, Invitrogen, Carlsbad, CA)
E. coli ER2738 host cells at 37?C for 5 h, after which the bacteria
were removed by centrifugation and the phage were precipitated
from the supernatant with polyethylene glycol (PEG)/NaCl
solution (20% PEG-8000, 2.5 M NaCl) (Fig. 1(c), step E). The
concentration of the amplified eluent was determined by phage
titering on LB X-gal (5-bromo-4-chloro-3-indolyl b-D-galactopyr-
and UV absorption spectroscopy.5We performed two rounds of
selection, and used the selected phage from each round to infect
plated ER2738 cells. We randomly picked individual plaques and
the selected clones were grown and purified for DNA sequencing
(MCLab, South San Francisco, CA) (Fig. 1(c), step D).
In the first round of selection, we fixed the buffer flow rate at
on the resulting peptide sequences. We tested four different washing
durations (5, 10, 30 and 60 min) and measured the percentage of
phage bound to the target (both specifically and non-specifically),
percent binding ¼eluted phage
input phage? 100%
We observed a non-linear, inversely proportional relationship
between the percentage of bound phage and the washing time, as the
The removal of the phage during the washing step can be modeled as
a first order process described by a simple exponential decay22,23
S ¼ S0e?kdt
where S is the density function at time t for the phage subpopulation
remaining on the target, S0is the amplitude constant and kdis the
dissociation rate constant. In the first round of selection, we obtained
a dissociation rate constant of kd1¼ 1.0 ? 0.1 ? 10?3s?1during
showing the channel design, nickel pattern and flow path. The device
dimensions are 64 mm ? 15.7 mm ? 1.5 mm (L ? W ? H). The height
and width of microfluidic channel are 30 mm and 12 mm, respectively. (b)
Bright field optical micrographs of nickel pattern in the microchannel.
Left: when an external field is applied, the large magnetic field gradients
at the edges of the nickel pattern effectively trap the beads. Right: when
the external field is removed, the nickel pattern is de-magnetized and the
beadsare efficientlyeluted. (c)Selection of the phagedisplay libraryusing
the MMS device. Step A: The phage library and the magnetic beads are
mixed and incubated for 30 min at room temperature with gentle agita-
tion. Step B: NeFeB permanent magnets are applied to the MMS device
to trap the magnetic beads carrying the phage bound to conjugated target
(a) Micrograph of the micromagnetic separation (MMS) device
proteins. The beads are held in place and washed under controlled
conditions. Then, the nickel patterns are de-magnetized by removing the
external magnets, and the phage-carrying beads are eluted. Step C: The
phage are dissociated from the target proteins by competitive elution.
Step D: Isolated phage are amplified via infection of E. coli cells, and
subsequently purified with PEG/NaCl solution. Step E: Clones from each
round of selection are randomly picked and their DNA is sequenced.
1034 | Lab Chip, 2009, 9, 1033–1036 This journal is ª The Royal Society of Chemistry 2009
In order to ensure that we retained most of the target-binding
clones during the first selection round, we adopted a two-tiered
stringency strategy,4,24wherein we used a low stringency selection
condition in the first round and high stringency selections in the
second round. The phage obtained after 5 min washing in the first
round (Fig. 2(a)) were subjected to a second round of selection with
ml/h. Again, we observed a non-linear, inverse correlation between
the percentage of bound phage and the washing duration (Fig. 2(b)).
However, the percentage of target-binding phage was significantly
higher in the second round than in the first. For example, in the first
round, 0.004% of the library was bound after 5 min of washing,
whereas 0.03% was bound after the second round. Interestingly, the
dissociation rate constant measured in the second round was kd2¼
1.07 ? 0.04 ? 10?3s?1which closely matched that of the first round.
Thus, we hypothesize that the dissociation rate constant of non-
specifically bound phage are similar for both rounds.
In addition, we noted that the frequency of occurrence of the well-
known SA binding peptide motif (histidine–proline–glutamine,
remarkable that, after 120 minutes of washing, 8 out of 9 sequenced
clones exhibited the HPQ motif (Fig. 2(b)), demonstrating the
importance of controlling the washing stringency. It is also note-
worthy that the probability of obtaining consensus sequences did not
converge to 100% even at the highest washing stringency; we suspect
that this may be caused by non-specific binding of the phage to the
beads and/or the MMS device.
To investigate the effect of flow rates on the resulting peptide
sequences, phage obtained after 5 min washing in the first round
(Fig. 2(a)) were utilized again in a second round where the washing
duration was fixed at 30 minutes, but the flow rates were varied (1, 5,
in the percentage of bound-phage as a function of increasing flow
flow rates. As the flow rate was increased from 1 ml/h to 20 ml/h, the
frequency of obtaining HPQ consensus monotonically increased
from 0% (0/10) to 78% (7/9) (Fig. 2(c)).
In conclusion, we have demonstrated the use of microfluidic
technology to enable precise control over washing stringency during
phage selection, which has a critical impact on the quantity and
diversity of the resulting target-binding peptide sequences. We have
also shown that both the wash duration and the flow rate can be
tailored to rapidly isolate high affinity binders. Though this was not
demonstrated here, our MMS device could be easily adapted to
control the stringency of other factors that influence selection
selection. (a) In the first round of selection, the percentage of recovered
phage as a function of washing time decays non-linearly, as non-specif-
ically bound and weak binding phage are removed. When modeled as
a first order exponential (dashed line), the dissociation rate constant was
kd1¼ 1.0 ? 0.1 ? 10?3s?1. (Inset) The canonical target binding peptide
The importance of controlling washing conditions during phage
motif (HPQ)was notfoundin the firstround.(b)In the secondround,the
percentage of the bound phage also showed an exponential decay as
stringency (washing time) increased, with a remarkably similar dissoci-
ation rate constant of kd2¼ 1.07 ? 0.04 ? 10?3s?1. (Inset) The percentage
of clones with the HPQ motif increased monotonically as a function of
washing time; After 120 minutes of washing, 8 out of 9 clones contained
the HPQ motif. (c) The percentage of bound phage as a function of flow
rate also showed an inversely proportional, non-linear relationship.
(Inset) The percentage of clones displaying the HPQ motif increased
monotonically as the flow rate increased, and after 30 minutes at the
highest flow rate (20 ml/h), 7 out of 9 clones displayed the HPQ motif. All
measurements were taken in triplicate.
This journal is ª The Royal Society of Chemistry 2009Lab Chip, 2009, 9, 1033–1036 | 1035
stringency, such as target concentration, detergent content, pH, salt Download full-text
concentration and temperature. Finally, our device architecture
readily lends itself towards incorporation into an integrated system
for automated directed evolution. In such a system, sample mixing,
incubation and washing26could be combined on the chip, potentially
offering higher degrees of accuracy, precision and reproducibility in
applying selection pressures that may not be possible with conven-
tional macroscopic approaches.
We thank the financial support from Office of Naval Research
(N00014-08-1-0469), National Institutes of Health (447850-24360)
and Armed Forces Institute for Regenerative Medicine (447852-
59819). YL is grateful for the postdoctoral fellowship support from
the California Institute of Regenerative Medicine (CIRM). KT
Collaborative Biotechnologies (DAAD1903D004). We acknowledge
B. Scott Ferguson, Jiangrong Qian and Yi Xiao at UCSB for their
invaluable assistance, and Yanting Zhang, Paul Pagano, David
Chang-Yen, Nancy Stagliano, and Andre’ DeFusco at Cynvenio
Biosystems (Santa Barbara, CA) for helpful discussions. Micro-
fabrication was carried out in the Nanofabrication Facility at UC
Notes and references
1 G. P. Smith, Science, 1985, 228, 1315–1317.
2 J. K. Scott and G. P. Smith, Science, 1990, 249, 386–390.
3 A. Sergeeva, M. G. Kolonin, J. J. Molldrem, R. Pasqualini and
W. Arap, Advanced Drug Delivery Reviews, 2006, 58, 1622–1654.
4 G. P. Smith and V. A. Petrenko,Chemical Reviews, 1997, 97, 391–410.
5 C. F. Barbas, D. R. Burton, J. K. Scott and G. J. Silverman, Phage
Display: A Laboratory Manual, Cold Spring Harbor Laboratory,
Cold Spring Harbor, 2001.
6 K. Nord, O. Nord, M. Uhlen, B. Kelley, C. Ljungqvist and
P. A. Nygren, European Journal of Biochemistry, 2001, 268, 4269–4277.
7 A. Biroccio, J. Hamm, I. Incitti, R. De Francesco and L. Tomei,
Journal of Virology, 2002, 76, 3688–3696.
8 D. Legendre, P. Soumillion and J. Fastrez, Nature Biotechnology,
1999, 17, 67–72.
9 D. Lu, J. Q. Shen, M. D. Vil, H. F. Zhang, X. Jimenez, P. Bohlen,
L. Witte and Z. P. Zhu, Journal of Biological Chemistry, 2003, 278,
10 A. Menendez and J. K. Scott, Analytical Biochemistry, 2005, 336,
11 G. Q. Zhuang, Y. Katakura, T. Furuta, T. Omasa, M. Kishimoto and
K. Suga, Journal of Bioscience and Bioengineering, 2001, 91, 474–481.
12 J. Persson, P. Augustsson, T. Laurell and M. Ohlin, Febs Journal,
2008, 275, 5657–5666.
13 P. H. Bessette, X. Y. Hu, H. T. Soh and P. S. Daugherty, Analytical
Chemistry, 2007, 79, 2174–2178.
14 U. Kim, C. W. Shu, K. Y. Dane, P. S. Daugherty, J. Y. J. Wang and
H. T. Soh, Proceedings of the National Academy of Sciences of the
United States of America, 2007, 104, 20708–20712.
15 M. S. Pommer, Y. T. Zhang, N. Keerthi, D. Chen, J. A. Thomson,
C. D. Meinhart and H. T. Soh, Electrophoresis, 2008, 29, 1213–1218.
16 A. Visser, B. H. Kunst, H. Keller and A. Schots, Current
Pharmaceutical Biotechnology, 2004, 5, 173–179.
17 P. S. Dittrich and A. Manz, Nature Reviews Drug Discovery, 2006, 5,
18 J. Qian, X. Lou, Y. Zhang, Y. Xiao, A. Gerdon and H. T. Soh, in
The Proceedings of MicroTAS 2008 Conference, San Diego, CA,
2008, pp. 1450–1452.
19 A. E. Gerdon, J. Qian, Y. Zhang, J. D. Adams, S. Oh, A. Csordas and
H. T. Soh, in The Proceedings of MicroTAS 2008 Conference, San
Diego, 2008, pp. 1594–1596.
20 D. W. Inglis, R. Riehn, R. H. Austin and J. C. Sturm, Applied Physics
Letters, 2004, 85, 5093–5095.
21 J. D. Adams, U. Kim and H. T. Soh, Proceedings of the National
Academy of Sciences of the United States of America, 2008, 105,
22 W. Mandecki, Y. C. J. Chen and N. Grihalde, Journal of Theoretical
Biology, 1995, 176, 523–530.
23 D. J. Oshannessy, M. Brighamburke, K. K. Soneson, P. Hensley and
I. Brooks, Analytical Biochemistry, 1993, 212, 457–468.
24 B. Levitan, Journal of Molecular Biology, 1998, 277, 893–916.
25 L. B. Giebel, R. T. Cass, D. L. Milligan, D. C. Young, R. Arze and
C. R. Johnson, Biochemistry, 1995, 34, 15430–15435.
26 R. K. Mosing and M. T. Bowser, Journal of Separation Science, 2007,
1036 | Lab Chip, 2009, 9, 1033–1036 This journal is ª The Royal Society of Chemistry 2009