Development of an In Vitro Compartmentalization
Screen for High-Throughput Directed Evolution of [FeFe]
Hydrogenases
James A. Stapleton1, James R. Swartz1,2*
1Department of Chemical Engineering, Stanford University, Stanford, California, United States of America, 2Department of Bioengineering, Stanford University, Stanford,
California, United States of America
Abstract
Background: [FeFe] hydrogenase enzymes catalyze the formation and dissociation of molecular hydrogen with the help of a
complex prosthetic group composed of common elements. The development of energy conversion technologies based on
these renewable catalysts has been hindered by their extreme oxygen sensitivity. Attempts to improve the enzymes by
directed evolution have failed for want of a screening platform capable of throughputs high enough to adequately sample
heavily mutated DNA libraries. In vitro compartmentalization (IVC) is a powerful method capable of screening for multiple-
turnover enzymatic activity at very high throughputs. Recent advances have allowed [FeFe] hydrogenases to be expressed
and activated in the cell-free protein synthesis reactions on which IVC is based; however, IVC is a demanding technique with
which many enzymes have proven incompatible.
Methodology/Principal Findings: Here we describe an extremely high-throughput IVC screen for oxygen-tolerant [FeFe]
hydrogenases. We demonstrate that the [FeFe] hydrogenase CpI can be expressed and activated within emulsion droplets,
and identify a fluorogenic substrate that links activity after oxygen exposure to the generation of a fluorescent signal. We
present a screening protocol in which attachment of mutant genes and the proteins they encode to the surfaces of
microbeads is followed by three separate emulsion steps for amplification, expression, and evaluation of hydrogenase
mutants. We show that beads displaying active hydrogenase can be isolated by fluorescence-activated cell-sorting, and we
use the method to enrich such beads from a mock library.
Conclusions/Significance: [FeFe] hydrogenases are the most complex enzymes to be produced by cell-free protein
synthesis, and the most challenging targets to which IVC has yet been applied. The technique described here is an enabling
step towards the development of biocatalysts for a biological hydrogen economy.
Citation: Stapleton JA, Swartz JR (2010) Development of an In Vitro Compartmentalization Screen for High-Throughput Directed Evolution of [FeFe]
Hydrogenases. PLoS ONE 5(12): e15275. doi:10.1371/journal.pone.0015275
Editor: Floyd Romesberg, The Scripps Research Institute, United States of America
Received September 23, 2010; Accepted November 3, 2010; Published December 6, 2010
Copyright:  2010 Stapleton, Swartz. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was funded by the Global Climate and Energy Project (GCEP) at Stanford University (http://gcep.stanford.edu/), as well as a National Defense
Science and Engineering (NDSEG) Fellowship (http://ndseg.asee.org/) to JAS. The funders had no role in study design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: jswartz@stanford.edu
Introduction
[FeFe] hydrogenase enzymes are very active hydrogen produc-
ers [1] but are extremely sensitive to oxygen, which is thought to
diffuse through two putative gas channels in the protein to poison
the H-cluster cofactor at the active site [2]. This sensitivity reduces
the applicability of the enzymes in biotechnological hydrogen
production schemes, for which they are otherwise very promising.
Narrowing the gas channels may prevent oxygen from diffusing to
the active site, but finding mutations that accomplish this is a
difficult challenge. The failure of previous attempts at evolving
oxygen tolerance suggests that multiple synergistic mutations may
be required before any improvement is observed [3].
In vitro compartmentalization (IVC) is a technology with the
potential to enable high-throughput screening of [FeFe] hydrog-
enase mutants. In IVC, extremely small aqueous droplets
suspended in a continuous oil phase isolate individual mutant
DNA molecules, forming independent emulsion cell-free protein
synthesis (eCFPS) reactors. Analogous to cells in an in vivo screen,
the droplets co-localize the gene, the mutant protein it encodes,
and the products of the desired enzymatic activity [4]. Like other
in vitro methods such as ribosome display [5] and mRNA display
[6], IVC can accommodate very large mutant libraries and is free
of the biases inherent in in vivo platforms. However, IVC is unique
among high-throughput in vitro methods in its ability to screen for
multiple-turnover catalytic activity [7]. Droplet-based technology
is advancing rapidly as its potential for evaluating mutants [8],
determining the effects of drug candidates on individual
encapsulated cells [9,10], and accelerating DNA sequencing
[11,12] becomes apparent. Combining IVC with microfluidic
technology allows monodisperse emulsion droplets to be formed
[13], mixed [14], split [15], merged [10], incubated, thermocycled
[16], ordered, assayed for fluorescence [17], and sorted [18], all
within the confines of a small chip.
PLoS ONE | www.plosone.org 1 December 2010 | Volume 5 | Issue 12 | e15275

Depending on the target of the directed evolution project, IVC
can be configured as a selection (in which the mutant gene itself is
generally the substrate for the desired activity or binding) or as a
high-throughput screen in which fluorescence-activated cell
sorting (FACS) is used to analyze and sort microbeads [8,19] or
water-in-oil-in-water (w/o/w) double emulsions [20,21] on the
basis of fluorescence linked to the desired activity. The power of
FACS in directed evolution applications has previously been
demonstrated by techniques such as yeast display [22] and
bacterial surface display [23]. In the microbead display IVC
method, mutant DNA and the protein it encodes bind to the
surface of microbeads within emulsion droplets. The compart-
mentalization imposed by the droplets ensures that each gene and
its encoded protein bind to the same bead. The resultant physical
genotype-phenotype linkage is maintained following emulsion
breakage and bead pooling. If the desired enzymatic activity
generates a fluorescent product which can also bind to the surface
of the beads, the beads can be sorted by FACS following recovery
from the emulsion. Genes encoding positive mutants are then
amplified from the sorted beads by PCR. Attachment of
fluorescent products to beads has been achieved by caged-
biotinylation of the fluorogenic substrate [24] or by generation
of a radical that reacts with bead-bound proteins [19].
Recently, our laboratory demonstrated production of active
[FeFe] hydrogenases in the cell-free protein synthesis reactions at
the heart of IVC [25,26], making the development of an IVC
screen for oxygen-tolerant [FeFe] hydrogenases possible. Here we
apply IVC to the directed evolution of [FeFe] hydrogenase I
(referred to as CpI) from C. pasteurianum, a complex metalloenzyme
containing multiple iron-sulfur electron transport clusters and an
H-cluster catalytic prosthetic group composed of [4Fe-4S] and
[2Fe-2S] clusters augmented by cyanide, carbon monoxide, and
dithiolate ligands [26,27]. We demonstrate an IVC system that
links [FeFe] hydrogenase activity to the generation of a fluorescent
signal on the surface of microbeads. Cell-free protein synthesis
reactions within monodisperse emulsion droplets produce mutated
CpI proteins, which bind to the surfaces of beads along with their
encoding DNA templates. Following oxygen exposure, surviving
[FeFe] hydrogenase mutants reduce the fluorogenic compound
C12-resazurin, allowing detection by FACS. We demonstrate the
system by enriching a mock library for beads bound to CpI DNA.
Results and Discussion
[FeFe] hydrogenase can be produced by emulsion CFPS
Previously we demonstrated the ability to produce and activate
[FeFe] hydrogenases in bulk CFPS reactions. However, many
proteins have been found to be incompatible with CFPS
expression within emulsion droplets [28]. To test whether the
helper protein-mediated synthesis and installation of the H-cluster
prosthetic group could occur within emulsions, we added linear
DNA templates encoding CpI to extracts of E. coli BL21 DE3
containing the three [FeFe] hydrogenase maturases [29] from
Shewanella oneidensis, and emulsified the mixtures by stirring,
vortexing, or extruding the aqueous phase into various oil/
surfactant mixtures. Methyl viologen assays [30] detected
hydrogenase activity in the collected products of the eCFPS
reactions, indicating successful synthesis and activation of [FeFe]
hydrogenases within the emulsion compartments.
Resazurin derivatives as fluorescent sensors of
hydrogenase activity
Isolation of oxygen-tolerant mutants by FACS requires that a
fluorescent signal develop in the presence of an active hydrogenase
surviving oxygen exposure. We evaluated the ability of a number
of redox-sensitive fluorophores to exchange electrons directly with
hydrogenase. The only compatible fluorogenic molecule we
identified was resazurin (Figure 1a, also known as AlamarBlue),
which was irreversibly reduced to the fluorescent resorufin
(Figure 1b) by electrons liberated from hydrogen gas by CpI.
Resazurin and resorufin are not suitable for flow cytometry
because they quickly leak out of cells, and examination by
fluorescence microscopy revealed that they also leaked out of our
emulsion droplets. Fortunately, the resazurin derivative C12-
resazurin (Figure 1c), which has the same fluorescent properties as
resazurin, is much better retained by cells, and is marketed for use in
FACS. We tested C12-resazurin and found that it was retained by
our emulsion droplets much better than unmodified resazurin.
C12-resazurin is marketed as a cell viability assay reagent
because active metabolic pathways reduce it to the fluorescent
form. We found that it is also readily reduced by pathways active
in our E. coli cell extract. Co-location of the cell extract and C12-
resazurin within the same emulsion droplet would therefore result
in indiscriminate production of the fluorescent product. To avoid
this problem we adopted the microbead display strategy discussed
earlier, which creates a physical genotype-phenotype linkage that
allows the beads to be removed from emulsions, washed,
resuspended in a new assay solution, and re-emulsified. Strepta-
vidin-coated beads displaying biotinylated DNA templates and
biotinylated anti-hemagglutinin (HA) tag antibodies were mixed
into a CFPS reaction solution and emulsified into a continuous oil
phase. Within the droplets, triply HA-tagged hydrogenase was
synthesized, matured, and bound to the antibodies. The beads
could then be broken out of the emulsion, washed, and re-
emulsified for a separate fluorescence-generation reaction step. We
delivered C12-resazurin to pre-formed emulsion droplets [31] to
ensure that the reaction did not begin before the beads were
Figure 1. Resazurin and its derivatives. Resazurin (A) is irreversibly
reduced to fluorescent resorufin (B) by electrons liberated from
hydrogen gas by [FeFe] hydrogenase. C12-resazurin (C) is identical to
resazurin except for an additional hydrophobic twelve-carbon tail.
doi:10.1371/journal.pone.0015275.g001
IVC Screen for Directed Evolution of Hydrogenases
PLoS ONE | www.plosone.org 2 December 2010 | Volume 5 | Issue 12 | e15275

compartmentalized. A schematic of the screening process is shown
in Figure 2a.
C12-resorufin nonspecifically adsorbs to the surfaces of the
polystyrene beads, allowing beads recovered from fluorescence-
generation emulsions to be directly sorted by FACS. Beads coated
with C12-resorufin can be washed and stored for weeks with little
decay of the fluorescent signal. Switching from the 1 mm diameter
beads used in previous reports to 5.6 mm diameter beads increased
the surface area available to bind C12-resorufin, and dramatically
improved signal resolution.
Hydrogenase-bound beads can be sorted by
C12-resorufin fluorescence
To test the feasibility of sorting beads on the basis of C12-resorufin
fluorescence, we mixed beads coated with anti-HA antibodies and
,1000 molecules of CpI DNA per bead with a 50:1 excess of beads
coated with antibodies but no DNA. The bead mixture was
suspended in CFPS reaction solution, emulsified into oil, and
incubated. Following protein synthesis the beads were recovered,
resuspended in assay solution, re-emulsified, and incubated to allow
antibody-bound hydrogenase to reduce C12-resazurin to C12-
resorufin. Finally, the beads were recovered and analyzed for C12-
resorufin fluorescence by FACS. In the fluorescence histogram of
the single sample, two separate populations are easily distinguished
(Figure 3a). The result confirms that the fluorescent C12-resorufin
molecules were unable to leak between emulsion compartments
during the fluorescence generation incubation, and that C12-
resorufin fluorescence generated by hydrogenase activity can be
used to identify active enzymes.
To confirm that exposure of wild-type hydrogenase to oxygen
would decrease subsequent fluorescence generation, we incubated
hydrogenase-coated beads in an air-equilibrated buffer solution.
As expected, following oxygen exposure the ability of the beads to
generate C12-resorufin fluorescence was diminished (Figure 3b).
A microfluidic chip allows the creation of monodisperse
emulsion droplets
Emulsion droplets produced by bulk methods such as stirring
are highly polydisperse (Figure 4a), potentially leading to uneven
protein production across otherwise identical beads. To improve
Figure 2. A schematic representation of the IVC screen protocol. The lower section of the diagram shows the molecules bound to the surface
of the streptavidin-coated beads at each step. Blue rectangles represent template DNA, green split arrows represent biotinylated anti-HA antibodies,
turquoise clouds represent 3xHA-tagged hydrogenase proteins, blue and magenta circles represent C12-resazurin and C12-resorufin. A. DNA-bound
beads are incubated with biotinylated anti-HA antibodies, added to a cell-free protein synthesis mixture, and emulsified into an oil phase. Emulsion
CFPS transcribes mRNA from the bound DNA templates and synthesizes HA-tagged hydrogenase proteins, which are bound by the antibodies. The
beads are then recovered from the emulsion, washed, and exposed to oxygen to challenge the hydrogenase mutants. The beads are then re-
emulsified, and C12-resazurin is delivered to the emulsion droplets. Mutant hydrogenases that survive oxygen exposure consume hydrogen and
reduce C12-resazurin to fluorescent C12-resorufin, which adsorbs to the bead surface. The beads are recovered and sorted by FACS. Fluorescent beads
are added to a PCR mixture and thermocycled to recover DNA encoding improved hydrogenase mutants. B. An emulsion PCR step amplifies unique
mutant templates to amounts sufficient to result in strong fluorescent signals. Beads are first incubated with biotinylated primers (represented by
short black lines) and less than one molecule of template DNA per bead, then added to a PCR mixture and emulsified. The emulsion is thermocycled,
and the beads are recovered, washed, and screened with the procedure shown in A.
doi:10.1371/journal.pone.0015275.g002
IVC Screen for Directed Evolution of Hydrogenases
PLoS ONE | www.plosone.org 3 December 2010 | Volume 5 | Issue 12 | e15275

droplet monodispersity, we designed a flow-focusing microfluidic
chip [32] capable of forming droplets at rates of several kilohertz
(Figure 4c). The design of the chip was based on published
microfluidic droplet generators. Two aqueous solutions, one
consisting of the beads and the small molecule components of
the eCFPS reaction mixture and the other primarily of diluted cell
extract, were injected into the chip through separate ports. The
solutions met immediately upstream of the emulsification nozzle,
ensuring that transcription did not begin before the beads were
compartmentalized within droplets. Experiments in which fluo-
rescein was added to one of the two mixtures confirmed that the
solutions evenly mixed within the droplets (Figure 4b), providing a
complete and uniform eCFPS reaction mixture. The throughput
of the chip, while much lower than that of bulk emulsification
methods, is comparable to that of FACS, and therefore does not
dramatically decrease the throughput of the overall process.
Emulsion PCR amplifies a single template molecule to
improve eCFPS yield
Though we observed a strong, saturated fluorescent signal when
we attached ,1000 DNA molecules to each bead, when we
attached single DNA templates to the beads we found that the
reduced numbers of enzyme copies that resulted were insufficient to
generate significant fluorescence. Successful microbead display
screens reported in the literature have incorporated mechanisms,
such as a caged biotin, to keep the fluorogenic substrate soluble
during the reaction period and only allow it to bind to the bead once
the reaction is complete. In our system, C12-resazurin presumably
binds to the bead immediately upon delivery into the aqueous
compartment. We hypothesized that the bound hydrogenase
enzymes were only able to reduce C12-resazurin molecules bound
nearby on the bead surface, which neither exchanged with C12-
resorufin in solution nor diffused along the surface of the bead.
To solve this problem, we amplified the single template copy
using emulsion PCR (ePCR, Figure 2b), a technique which has
recently seen extensive use in next-generation sequencing [9,33,34],
genotyping [35], and directed evolution [36]. Biotinylated primers
were bound to the beads in numbers small enough to preserve
ample streptavidin binding sites for the downstream antibody
binding step. Amplification by ePCR of the single DNA molecule
attached to the bead increased the number of copies of template in
the subsequent eCFPS reaction, thereby increasing hydrogenase
production and eventual signal generation. Compartmentalization
of the beads within emulsion droplets is necessary during the ePCR
step to ensure that each bead only binds DNA amplified from the
single template on its surface, and does not exchange DNA with
other beads. Only after the addition of the ePCR step to the
protocol were we able to measure fluorescent signals from beads
originally bound to single molecules of DNA.
Since amplification efficiency is dependent on droplet size, we
formed emulsions for ePCR in the microfluidic device. The
amplification efficiency of ePCR drops markedly with increasing
amplicon length [37], and very low yields have been reported for
targets the size of the CpI expression template (,2 kb). Still,
template dilution experiments indicated that even a modest increase
in the number of DNA copies displayed on the bead could allow a
significant increase in protein expression. Because amplification of
long templates has been shown to be more efficient in larger
emulsion droplet volumes [38], we set the oil and aqueous phase
flow rates to produce droplets about 30 microns in diameter.
An additional benefit of the ePCR step is improved efficiency of
DNA recovery by PCR from sorted beads following FACS. Some
fraction of the bead-bound genes are expected to be partially or
completely degraded by nucleases during the eCFPS incubation.
Figure 3. Beads can be sorted on the basis of C12-resorufin
fluorescence linked to hydrogenase activity. A. The distinct peak
separation between the two fluorescence populations after FACS
analysis of a sample consisting of a mixture of beads with and without
bound hydrogenase confirms that fluorescent C12-resorufin does not
leak between emulsion droplets. The antibody-coated beads were
incubated with a bulk CFPS reaction mixture synthesizing triply HA-
tagged hydrogenase or with a template-free reaction mixture. The
beads were then washed, mixed at a 1:50 ratio, and subjected to the
fluorescence generation emulsion step and subsequent FACS analysis.
B. Decrease in bead fluorescence generation ability caused by
deactivation of hydrogenase following incubation in a buffer containing
,0.25 mM dissolved oxygen. Following hydrogenase production by
eCFPS, the beads were recovered and washed. The sample was split in
half, and one half was incubated in 1.5 mL of air-equilibrated 50 mM
Tris-HCl pH 9 for 60 minutes. Both samples were then subjected to the
fluorescence generation emulsion step and evaluated by FACS.
doi:10.1371/journal.pone.0015275.g003
IVC Screen for Directed Evolution of Hydrogenases
PLoS ONE | www.plosone.org 4 December 2010 | Volume 5 | Issue 12 | e15275

Following ePCR, it should be much more likely that at least one
copy of each gene will remain competent for amplification.
The microbead display format adopted here is a very useful
option for FACS analysis of IVC-generated fluorescent signals.
However, caged-biotinylated fluorogenic substrates or other
convenient methods to control fluorophore/bead binding are
often unavailable. It thus seems likely that an ePCR step will prove
useful in many future applications of IVC.
Enrichment of beads binding hydrogenase DNA
To demonstrate the ability of the screen to identify beads bound
to single molecules of hydrogenase DNA, we performed a test
enrichment experiment. We incubated two sets of beads with
single-molecule per bead levels of biotinylated DNA templates
encoding either tagged CpI or a negative control protein,
chloramphenicol acetyltransferase (CAT). Both templates con-
tained the same homoprimer [39] annealing sites for PCR
amplification. We mixed the CAT and CpI beads at a 20:1 ratio,
performed the three emulsification steps of the IVC screen
protocol (omitting the oxygen exposure step), and sorted the
recovered beads by FACS. Low- and high-fluorescence beads were
sorted and collected in separate tubes.
Amplification of the DNA on the sorted low-fluorescence beads
gave rise to a bright CAT gene band and a very light CpI gene
band, while the high-fluorescence beads yielded a bright CpI gene
band and a light CAT gene band (Figure 5). This enrichment of
CpI DNA indicates that the IVC screen was able to identify and
sort beads initially bound to single molecules of DNA encoding an
active [FeFe] hydrogenase. No enrichment was observed when the
ePCR step was omitted from the protocol.
The success of the enrichment experiment indicates that [FeFe]
hydrogenase mutants can be screened in an extremely high-
throughput fashion by in vitro compartmentalization using C12-
resazurin to link enzyme activity to the development of a florescent
signal. However, further work will be required before screening of
true mutant libraries can begin. The enrichment experiment
described above used two populations of beads that were incubated
with DNA separately and then mixed. A more realistic scenario is
one in which the DNA is mixed before incubation with beads.
Enrichment of hydrogenase DNA from mixtures with excesses of
decoy DNA much larger than 20:1 must also be demonstrated.
We also tested the platform described here with a second [FeFe]
hydrogenase, HydA1 from Chlamydomonas reinhardtii, and confirmed
its compatibility with both eCFPS and C12-resazurin. While this
IVC screen was developed with the goal of discovering oxygen-
tolerant hydrogenase mutants, it is actually a screen for hydrogenase
activity, useful for screening for mutants with any of a number of
industrially important characteristics, such as thermostability (by
incubating beads at high temperatures between the eCFPS and C12-
resazurin emulsion steps), tolerance for conditions such as might
exist in a fuel cell, and so on. Hydrogenases with any of these
phenotypes can be screened for with only minimal modifications to
the platform described here. The throughput of IVC (thousands of
beads per second on modern FACS machines) represents a
dramatic improvement over traditional 96-well microtiter plate-
based screens, and the methods described here provide a way
forward towards the discovery of hydrogenase mutants with the
potential to revolutionize the world’s energy economy.
Materials and Methods
DNA preparation
The DNA template encoding C-terminally triply HA-tagged
templates was created by overlap extension PCR using the pK7-
Figure 4. Creation of monodisperse emulsion droplets with a
microfluidic device. A. A polydisperse water-in-oil emulsion created
by stirring. Beads (5.6 mm, orange circle) encapsulated in droplets are
visible. B. Monodisperse droplets produced with the microfluidic
device. Two solution streams, one containing 5 mM fluorescein, met
immediately upstream of the emulsification nozzle. The resulting
droplets were evenly fluorescent, indicating that each contained an
equal mixture of the two solutions and that the laminar solutions mixed
within the droplets. Beads (5.6 mm, orange circle) are visible within the
droplets. C. Design of a microfluidic chip for the production of
monodisperse emulsion droplets. Circular structures indicate punch
holes for insertion of the steel pins through which fluids enter and exit
the chip. Aqueous phases enter on the left, oil enters on the right, and
the emulsified product exits from the punch hole in the center.
Microfluidic filter pillars prevent dust and other obstructions from
clogging the nozzle. Main channels are 100 mm wide, and narrow to
20 mm. All features are 25 mm in height. Inset photograph shows the
nozzle during the formation of the droplets pictured in B.
doi:10.1371/journal.pone.0015275.g004
IVC Screen for Directed Evolution of Hydrogenases
PLoS ONE | www.plosone.org 5 December 2010 | Volume 5 | Issue 12 | e15275

CpI plasmid [26] with the following primers, which annealed
outside the T7 RNA polymerase promoter and terminator and
contained extensions for the addition of the TR homoprimer
sequence [40] (59-CCATTCATTAATGCACATAACTAT-39)
and the 3xHA tag: 59-CCA TTC ATT AAT GCA CAT AAC
TAT TCC AAC GAG TTC GCG GCC GCT TAG GCA CCC
CAG GCT TTA C-39, 59-GGC ATA GTC TGG GAC GTC
ATA TGG ATA TGC ATA GTC CGG AAC ATC GTA TGG
GTA TTT TTT ATA TTT AAA GTG CAG GAT TTC-39, 59-
CCA TTC ATT AAT GCA CAT AAC TAT TCC AAC GAG
TTC GAC GAG CGT CAG CTT GCA TGC CCT GCA GCT-
39, 59-TAT CCA TAT GAC GTC CCA GAC TAT GCC TAC
CCG TAT GAC GTG CCA GAT TAC GCG TAA TAA TAA
TTT TTT TAA GGC AGT TAT TGG TG-39. The CAT linear
template was created from the pK7-CAT plasmid in the same
way. The linear templates were re-amplified with biotinylated TR
homoprimer to yield constructs with biotin groups at both ends.
Following purification with PCR cleanup kits (Qiagen, Valen-
cia, CA), DNA concentrations were measured with the Quant-It
dsDNA broad-range assay on a Qubit fluorimeter (Invitrogen,
Carlsbad, CA) following the manufacturer’s protocol. The
solutions were then diluted in water or TE buffer (10 mM Tris-
HCl pH 8, 1 mM EDTA) to 107 molecules/mL.
Bead preparation
26107 5.6 micron diameter streptavidin-coated beads (CP01N,
Bangs Laboratories, Fishers, IN) were washed with DNA binding
buffer (1 mM NaCl, 10 mM Tris-HCl, pH 7.5), resuspended in
the same, and incubated with 0.3–1 molecule of biotinylated linear
DNA template per bead for 16 hours at room temperature with
rotation to prevent settling. The beads were then incubated in
DNA binding buffer with 3 million biotinylated reverse primers
per bead for 2 hours, and washed in PBS.
Microfluidic chip design and preparation
A PDMS microfluidic flow-focusing emulsification device with
the design shown in Figure 4c was fabricated by the Stanford
Microfluidics Foundry (http://thebigone.stanford.edu/foundry/in-
dex.html). The chip was bonded to a glass substrate by PDMS-
PDMS bonding. The device was pre-treated with Aquapel [41]
(PPG Industries, Pittsburgh, PA) to render the channels hydropho-
bic and ensure that the oil phase preferentially wetted the surface.
The chip was designed with a uniform 25 mm feature height.
Channels narrow from a width of 100 mm to 20 mm at the nozzle.
Emulsion PCR
Following incubation with DNA templates and biotinylated
ePCR primers as described above, the beads were resuspended in
a mixture of 1X Pfu buffer, 0.2 mM each dNTP, 0.1 mM TR
homoprimer, 0.1% Tween 20, and 0.2 U/mL Pfu Turbo
polymerase (Stratagene, La Jolla, CA) at a bead concentration of
56104/mL.
The PCR mixture and an oil phase consisting of Fluorinert FC-
40 (Sigma Aldrich, St Louis, MO) with 2% w/w PFPE-PEG block
copolymer surfactant shown to be compatible with both PDMS
chips and eCFPS [42] (provided courtesy of Raindance Technol-
ogies, Lexington, MA) were forced into the inlets of the micro-
fluidic chip described above at constant flow rates with syringe
pumps. The second aqueous inlet was blocked with a clamped
tube. An aqueous flow rate of 6 mL/min and an oil flow rate of
50 mL/min generated monodisperse droplets approximately
30 mm in diameter.
The emulsion was distributed in 50 mL aliquots into thin-walled
PCR tubes and heated at 95uC for 1 minute, followed by 50 cycles of
95uC for 10 seconds, 55uC for 30 seconds, and 72uC for 2 minutes.
The thermocycled emulsions were collected into one tube and
broken by addition of a volume of A104 emulsion destabilizer
(provided by Raindance Technologies, Lexington, MA) equal to
10% of the volume of emulsified aqueous phase, followed by gentle
stirring with a pipette tip and recovery of the upper aqueous phase.
Beads were then incubated in PBS with 200,000 biotin-conjugated
rabbit anti-HA antibodies per bead (Immunology Consultants
Laboratory, Inc., Newberg, OR) for two hours at room temperature.
Cell-free Protein Synthesis
Inside an anaerobic glove box the cell-free extract harboring the
three [FeFe] hydrogenase helper proteins HydE, F, and G was
reconstituted [26] by incubation at room temperature with
0.5 mM ferrous ammonium sulfate and 0.5 mM Na2S, and then
spun at 10,000 g until all precipitates were removed. The
following two solutions were prepared (concentrations given are
those in the final CFPS mixture, which consisted of an equal
mixture of Solution A and B): Solution A consisted of a small
Figure 5. Enrichment of beads bound to single molecules of
hydrogenase DNA. Beads displaying CAT DNA and beads displaying
CpI DNA were originally mixed in a 20:1 ratio. Comparison of the PCR
amplification products from nonfluorescent (2) and fluorescent (+)
sorted beads by agarose gel electrophoresis indicates that the more
fluorescent population is enriched in CpI-bound beads. Nonspecific
amplification products are visible in both lanes.
doi:10.1371/journal.pone.0015275.g005
IVC Screen for Directed Evolution of Hydrogenases
PLoS ONE | www.plosone.org 6 December 2010 | Volume 5 | Issue 12 | e15275

molecule substrate mixture [25] and 0.1% Tween 20 (Sigma
Aldrich, St Louis, MO) to prevent bead aggregation. Solution B
consisted of 6.7 mg/mL bacteriophage lambda Gam protein (to
inhibit RecBCD exonuclease in the extract), 0.1 mg/mL T7 RNA
polymerase, and 25% v/v E. coli cell extract. Gam protein and T7
RNA polymerase were expressed and purified in house.
Solution A was added to washed, pelleted beads and vortexed to
give a suspension of 105 beads/mL. Solutions A and B were loaded
into separate Hamilton gastight syringes. The FC-40/surfactant
oil phase described above, having been stored in the glove box to
remove oxygen, was loaded into a third, larger syringe.
The microfluidic chip described above was used to generate
monodisperse emulsions. Flow rates of 3 mL/min of each of the
two aqueous phases and 50 mL/min of the oil phase produced
monodisperse droplets approximately 30 mm in diameter. Emul-
sification was carried out outside the glove box on an inverted
microscope. The emulsified mixture was collected into an airtight
vial (containing 500 mL buffer to dilute and thus prevent protein
synthesis from taking place in any cell-free mixture that avoids
emulsification) and returned to the glove box immediately upon
completion. The emulsified reactions were incubated at room
temperature for 6–16 hours. Following this incubation the
emulsion was broken with A104 emulsion destabilizer as before.
When emulsions were formed by bulk methods, 60 mL of the
combined CFPS mixture above (omitting Tween 20) was added to
the pelleted beads, mixed, and immediately emulsified into 500 mL
cold microbiology grade light mineral oil containing 1.8% v/v Span
80 and 0.2% v/v Triton X-100 (all from Sigma Aldrich, St Louis,
MO) surfactants. Emulsions were formed by stirring the mixture in
the bottom of a 2 mL round-bottom cryovial with a 3 mm68 mm
magnetic stir bar at 600 rpm for 7 minutes, briefly vortexing at
maximum speed after each minute. Alternatively, eCPFS emulsions
were prepared by extrusion, with an oil phase of mineral oil with
1.8% v/v Span 80 and 0.2% v/v Triton X-100, or of decane with
1% w/v Span 60 and 1% w/v cholesterol (all from Sigma Aldrich,
St Louis, MO). The beads were suspended in 50 mL of CFPS
solution and extruded through a 19 mm track-etch polycarbonate
membrane with 14 mm pores (GE Osmonics, Minnetonka, MN)
into 200 mL of oil using a hand extruder (Avanti Polar Lipids,
Alabaster, AL). After a total of 15 passes through the membrane the
emulsion was dispensed into a 1.5 mL tube. The emulsified CFPS
reaction droplets were incubated at room temperature for 16 hours.
After the incubation, the emulsion was centrifuged to settle the
droplets to the bottom of the tube, and the clarified oil phase was
removed. 400 mL of 10 mM Tris-HCl pH 8 with 0.5% v/v Tween
20 (Sigma Aldrich, St Louis, MO) was added, and the emulsion was
centrifuged at 10000 g for 3 minutes to break the beads out of the
destabilized droplets, vortexed, and centrifuged for another three
minutes. This resulted in a pellet of beads at the bottom of the tube,
buffer in the middle, and a thick oil phase at the top. The pelleted
beads were aspirated from the bottom of the tube with a pipette and
washed in a new tube with 200 mL 50 mM Tris-HCl pH 9, then
resuspended in the same buffer.
Oxygen Exposure
For oxygen exposure experiments, a 1.5 mL tube completely
filled with 10 mM Tris-HCl pH 8 equilibrated with air was moved
into the anaerobic chamber. Suspended beads were added to the
tube and incubated for 60–90 minutes before being pelleted by
centrifugation and washed with anaerobic buffer.
Fluorescent signal generation
Oxygen-exposed beads were suspended in 30 mL of 50 mM Tris-
HCl pH 9 and re-emulsified by extrusion as described above into
300 mL of mineral oil mixed with 3% v/v Abil EM90 surfactant
(Evonik Degussa, Essen, Germany), which has been reported to
perform well in redox-sensitive emulsion applications [43]. 0.3 mL
of a 10 mM solution of C12-resazurin (Invitrogen, Carlsbad, CA) in
DMSO was added to the emulsion, which was vortexed for 30
seconds to deliver the C12-resazurin to the droplets. The emulsion
was then incubated anaerobically in the dark at room temperature
for 2–16 hours. Following the incubation, the emulsion was
removed from the anaerobic chamber and centrifuged at
20,000 g for 1 minute, and the clarified excess oil phase was
removed. 400 mL of 10 mM Tris-HCl pH 8 with 0.5% v/v Tween
20 (Sigma Aldrich, St Louis, MO) was added to disrupt the
emulsion, which was then centrifuged at 20,000 g for three minutes,
vortexed, and centrifuged for another three minutes. The pelleted
beads were aspirated from the bottom of the tube with a pipette and
washed in a new tube with 200 mL 50 mM Tris-HCl pH 9 (the pH
at which C12-resorufin is maximally fluorescent), then resuspended
in the same buffer and analyzed by FACS.
FACS analysis and sorting
Beads were analyzed and sorted using various flow cytometers
at the Stanford Shared FACS Facility. A dye laser tuned to
570 nm was used for excitation, and emission was collected
through a 605/40 nm filter. Beads were analyzed at rates of 500–
1000 events per second.
Recovery of mutant DNA by PCR
Unsorted and sorted beads were amplified with Phusion DNA
polymerase (Finnzymes, Espoo, Finland) using a single-molecule
PCR protocol described previously [25].
Note on proprietary reagents: it is the policy of Raindance
Technologies to provide commercially unavailable reagents to
academic researchers at no cost.
Acknowledgments
The authors would like to acknowledge the contributions of the Stanford
Shared FACS Facility, the Stanford Microfluidics Foundry, Raindance
Technologies, the Cliff Wang Lab at Stanford University, and Bertrand
Lui.
Author Contributions
Conceived and designed the experiments: JAS JRS. Performed the
experiments: JAS. Analyzed the data: JAS JRS. Wrote the paper: JAS JRS.
References
1. Adams MW (1990) The structure and mechanism of iron-hydrogenases.
Biochimica et Biophysica Acta 1020: 115–145.
2. Cohen J, Kim K, Posewitz M, Ghirardi ML, Schulten K, et al. (2005) Molecular
dynamics and experimental investigation of H2 and O2 diffusion in [Fe]-
hydrogenase. Biochem Soc Transact 33: 80–82.
3. Stapleton JA, Swartz JR (2010) A cell-free microtiter plate screen for improved
[FeFe] hydrogenases. PLoS ONE 5(5): e10554. DOI:10.1371/journal.
pone.0010554.
4. Tawfik DS, Griffiths AD (1998) Man-made cell-like compartments for molecular
evolution. Nature Biotechnology 16: 652–656.
5. Hanes J, Pluckthun A (1997) In vitro selection and evolution of functional proteins
by using ribosome display. Proc Natl Acad Sci U S A 94: 4937–4942.
6. Roberts RW, Szostak JW (1997) RNA-peptide fusions for the in vitro selection of
peptides and proteins. Proc Natl Acad Sci U S A 94: 12297–12302.
7. Agresti JJ, Kelly BT, Ja¨schke A, Griffiths AD (2005) Selection of ribozymes that
catalyse multiple-turnover Diels-Alder cycloadditions by using in vitro compart-
mentalization. Proc Natl Acad Sci U S A 102: 16170–16175.
8. Griffiths AD, Tawfik DS (2003) Directed evolution of an extremely fast
phosphotriesterase by in vitro compartmentalization. The EMBO Journal 22:
24–35.
IVC Screen for Directed Evolution of Hydrogenases
PLoS ONE | www.plosone.org 7 December 2010 | Volume 5 | Issue 12 | e15275

9. Aharoni A, Amitai G, Bernath K, Magdassi S, Tawfik DS (2005) High-
throughput screening of enzyme libraries: thiolactonases evolved by fluores-
cence-activated sorting of single cells in emulsion compartments. Chemistry &
Biology 12: 1281–1289.
10. Brouzes E, Medkova M, Savenelli N, Marran D, Twardowski M, et al. (2009)
Droplet microfluidic technology for single-cell high-throughput screening. Proc
Natl Acad Sci U S A 106: 14195–14200.
11. Kelly BT, Baret J, Taly V, Griffiths AD (2007) Miniaturizing chemistry and
biology in microdroplets. Chemical Communications DOI: 10.1039/b616252e.
12. Margulies M, Egholm M, Altman WE, Attiya S, Bader JS, et al. (2005) Genome
sequencing in microfabricated high-density picolitre reactors. Nature 437:
376–380.
13. Shah RK, Shum HC, Rowat AC, Lee D, Agresti JJ, et al. (2008) Designer
emulsions using microfluidics. Materials Today 11: 18–27.
14. Leamon JH, Link DR, Egholm M, Rothberg JM (2006) Overview: methods and
applications for droplet compartmentalization of biology. Nature Methods 3:
541–543.
15. Link D, Anna S,Weitz D, StoneH (2004) Geometrically mediated breakup of drops
in microfluidic devices. Physical Review Letters DOI: 10.1103/PhysRevLett.
92.054503.
16. Schaerli Y, Wootton RC, Robinson T, Stein V, Dunsby C, et al. (2009)
Continuous-flow polymerase chain reaction of single-copy DNA in microfluidic
microdroplets. Analytical Chemistry 81: 302–306.
17. Frenz L, Blouwolff J, Griffiths AD, Baret J (2008) Microfluidic production of
droplet pairs. Langmuir 24: 12073–12076.
18. Agresti JJ, Antipov E, Abate AR, Ahn K, Rowat AC, et al. (2010) Ultrahigh-
throughput screening in drop-based microfluidics for directed evolution. Proc
Natl Acad Sci U S A DOI: 10.1073/pnas.0910781107.
19. Sepp A, Tawfik DS, Griffiths AD (2002) Microbead display by in vitro
compartmentalisation: selection for binding using flow cytometry. FEBS letters
532: 455–458.
20. Mastrobattista E, Taly V, Chanudet E, Treacy P, Kelly BT, et al. (2005) High-
throughput screening of enzyme libraries: in vitro evolution of a beta-
galactosidase by fluorescence-activated sorting of double emulsions. Chemistry
& Biology 12: 1291–1300.
21. Bernath K, Hai M, Mastrobattista E, Griffiths AD, Magdassi S, et al. (2004) In
vitro compartmentalization by double emulsions: sorting and gene enrichment by
fluorescence activated cell sorting. Analytical Biochemistry 325: 151–157.
22. Boder ET, Wittrup KD (2000) Yeast surface display for directed evolution of
protein expression, affinity, and stability. Methods in Enzymology 328: 430–444.
23. Georgiou G, Stathopoulos C, Daugherty PS, Nayak AR, Iverson BL, et al.
(1997) Display of heterologous proteins on the surface of microorganisms: From
the screening of combinatorial libraries to live recombinant vaccines. Nature
Biotechnology 15: 29–34.
24. Pirrung MC, Huang CY (1996) A general method for the spatially defined
immobilization of biomolecules on glass surfaces using ‘‘caged’’ biotin.
Bioconjugate Chemistry 7: 317–321.
25. Boyer M, Stapleton JA, Kuchenreuther JM, Wang C, Swartz JR (2008) Cell-free
synthesis and maturation of hydrogenases. Biotechnology and Bioengineering
99: 59–67.
26. Kuchenreuther JM, Stapleton JA, Swartz JR (2009) Tyrosine, cysteine, and S-
adenosyl methionine stimulate in vitro [FeFe] hydrogenase activation. PLoS ONE
4(10): e7565. DOI:10.1371/journal.pone.0007565.
27. Peters JW, Lanzilotta WN, Lemon BJ, Seefeldt LC (1998) X-ray crystal structure
of the Fe-only hydrogenase (CpI) from Clostridium pasteurianum to 1.8 angstrom
resolution. Science 282: 1853–1858.
28. Davidson EA, Dlugosz PJ, Levy M, Ellington AD (2009) Directed evolution of
proteins in vitro using compartmentalization in emulsions. Current Protocols in
Molecular Biology DOI: 10.1002/0471142727.mb2406s87.
29. Posewitz MC, King PW, Smolinski SL, Zhang L, Seibert M, et al. (2004)
Discovery of two novel radical S-adenosylmethionine proteins required for the
assembly of an active [Fe] hydrogenase. The Journal of Biological Chemistry
279: 25711–20.
30. Peck H, Gest H (1956) A new procedure for assay of bacterial hydrogenases.
Journal of Bacteriology 71: 70–80.
31. Bernath K, Magdassi S, Tawfik DS (2005) Directed evolution of protein
inhibitors of DNA-nucleases by in vitro compartmentalization (IVC) and nano-
droplet delivery. Journal of Molecular Biology 345: 1015–1026.
32. Anna SL, Bontoux N, Stone HA (2003) Formation of dispersions using ‘‘flow
focusing’’ in microchannels. Applied Physics Letters 82: 364–366.
33. Shendure J, Porreca G, Reppas N, Lin X, McCutcheon J, et al. (2005) Accurate
multiplex polony sequencing of an evolved bacterial genome. Science 309:
1728–1732.
34. Pushkarev D, Neff NF, Quake SR (2009) Single-molecule sequencing of an
individual human genome. Nature Biotechnology DOI: 10.1038/nbt.1561.
35. Pohl G, Shih I (2004) Principle and applications of digital PCR. Expert Review
of Molecular Diagnostics 4: 41–47.
36. Gan R, Yamanaka Y, Kojima T, Nakano H (2008) Microbeads display of
proteins using emulsion PCR and cell-free protein synthesis. Biotechnology
Progress 24: 3–10.
37. Diehl F, Li M, He Y, Kinzler K, Vogelstein B (2006) BEAMing: single-molecule
PCR on microparticles in water-in-oil emulsions. Nature Methods 3: 551–559.
38. Kumaresan P, Yang CJ, Cronier SA, Blazej RG, Mathies RA (2008) High-
throughput single copy DNA amplification and cell analysis in engineered
nanoliter droplets. Analytical Chemistry 80: 3522–3529.
39. Brownie J, Shawcross S, Theaker J, Whitcombe D, Ferrie R, et al. (1997) The
elimination of primer-dimer accumulation in PCR. Nucleic Acids Research 25:
3235–3241.
40. Rungpragayphan S, Kawarasaki Y, Imaeda T, Kohda K, Nakano H, et al.
(2002) High-throughput, cloning-independent protein library construction by
combining single-molecule DNA amplification with in vitro expression. Journal of
Molecular Biology 318: 395–405.
41. Edd JF, Di Carlo D, Humphry KJ, Ko¨ster S, Weitz DA, et al. (2008) Controlled
encapsulation of single-cells into monodisperse picolitre drops. Lab on a Chip 8:
1262–1264.
42. Holtze C, Rowat AC, Agresti JJ, Hutchison JB, Angile` FE, et al. (2008)
Biocompatible surfactants for water-in-fluorocarbon emulsions. Lab on a Chip
8: 1632–1639.
43. Ghadessy FJ, Holliger P (2004) A novel emulsion mixture for in vitro
compartmentalization of transcription and translation in the rabbit reticulocyte
system. Protein Engineering, Design, and Selection 17: 201–204.
IVC Screen for Directed Evolution of Hydrogenases
PLoS ONE | www.plosone.org 8 December 2010 | Volume 5 | Issue 12 | e15275