ARTICLE
Compartmentalised RNA catalysis
in membrane-free coacervate protocells
Björn Drobot
1
, Juan M. Iglesias-Artola
1
, Kristian Le Vay
2
, Viktoria Mayr
2
, Mrityunjoy Kar
1
,
Moritz Kreysing
1
, Hannes Mutschler
2
& T-Y Dora Tang
1
Phase separation of mixtures of oppositely charged polymers provides a simple and direct
route to compartmentalisation via complex coacervation, which may have been important for
driving primitive reactions as part of the RNA world hypothesis. However, to date, RNA
catalysis has not been reconciled with coacervation. Here we demonstrate that RNA catalysis
is viable within coacervate microdroplets and further show that these membrane-free dro-
plets can selectively retain longer length RNAs while permitting transfer of lower molecular
weight oligonucleotides.
DOI: 10.1038/s41467-018-06072-w
OPEN
1
Max-Planck Institute for Molecular Cell Biology and Genetics, Pfotenhauerstraße 108, 01307 Dresden, Germany.
2
Max-Planck Institute for Biochemistry,
Am Klopferspitz 18, 82152 Martinsried, Germany. These authors contributed equally: Juan M. Iglesias-Artola, Kristian Le Vay. Correspondence and requests
for materials should be addressed to M.K. (email: kreysing@mpi-cbg.de) or to H.M. (email: mutschler@biochem.mpg.de)
or to T-Y.D.T. (email: tang@mpi-cbg.de)
NATURE COMMUNICATIONS | (2018) 9:3643 | DOI: 10.1038/s41467-018-06072-w | www.nature.com/naturecommunications 1
1234567890():,;
C
ompartmentalisation driven by spontaneous self-assembly
processes is crucial for spatial localisation and con-
centration of reactants in modern biology and may have
been important during the origin of life. One route known as
complex coacervation describes a complexation process
1,2
between two oppositely charged polymers such as polypeptides
and nucleotides
3–7
. The resulting coacervate microdroplets are
membrane free, chemically enriched and in dynamic equilibrium
with a polymer poor phase. In addition to being stable over a
broad range of physicochemical conditions, coacervate droplets
are able to spatially localise and up-concentrate different
molecules
3,8
and support biochemical reactions
9,10
.
It has been hypothesised that compartments which form via
coacervation could have played a crucial role during the origin of
life by concentrating molecules and thus initiating the first bio-
chemical reactions on Earth
11
. Coacervation has also been
implicated in modern biology where it has been shown that the
formation of membrane-free compartments or condensates such
as P-bodies or stress granules within cells are driven by this
mechanism
12,13
. These membrane-free organelles are chemically
isolated from their surrounding cytoplasm through a diffusive
phase boundary, permitting the exchange of molecules with their
surroundings
14
. In addition, these compartments may localise
specific biological reactions and play important roles in cellular
functions such as spatial and temporal RNA localisation within
the cell
15–18
.
Whilst there is increasing evidence for the functional impor-
tance of RNA compartmentalisation via coacervation in modern
biology, this phenomenon would also have been vitally important
during a more primitive biology. Up-concentration and locali-
sation could have enabled RNA to function both as a catalyst
(ribozyme) and storage medium for genetic information, as
required by the RNA world hypothesis
19
. To date, ribozymes have
been encapsulated within eutectic ice phases
20,21
and protocell
models such as water–oil-droplets for directed evolution experi-
ments
22–24
, membrane-bound lipid vesicles
25–27
, and membrane-
free compartments based on polyethelene glycol (PEG)/dextran
aqueous two-phase systems (ATPS)
28
. Interestingly, RNA cata-
lysis within ATPS exhibits an increased rate of reaction as a result
of the increased concentration within the dextran phase. Despite
these examples, RNA catalysis has not been demonstrated within
coacervate-based protocells. Therefore, coacervate protocells
based on carboxymethyl dextran sodium salt (CM-Dex) and
poly-
L-lysine (PLys) (Supplementary Fig. 1) were chosen as the
model system due to their proven ability to encapsulate and
support complex biochemical reactions catalysed by highly
evolved enzymes
10
. In contrast to these enzymes, structurally
simple ribozymes, which are thought to have played a key role
during early biology, are prone to fold into inactive conforma-
tions in the absence of RNA chaperones or additional auxiliary
elements
29–33
, and therefore may be rendered inactive by inter-
actions within the highly charged and crowded interior of coa-
cervate microdroplets. Herein, we directly probe the effect of the
coacervate microenvironment on primitive RNA catalysis, and
show the ability of the coacervate microenvironment to support
RNA catalysis whilst selectively sequestering ribozymes and
permitting transfer of lower molecular weight oligonucleotides.
Results
Hammerhead activity in bulk coacervate phase. We developed a
real-time fluorescence resonance energy transfer (FRET) assay
(see Methods) to investigate the effect of the coacervate micro-
environment on catalysis of a minimal version of the hammer-
head ribozyme derived from satellite RNA of tobacco ringspot
virus (HH-min)
34
. HH-min and its FRET substrate (Fig. 1a,
Methods) were incubated within a bulk polysaccharide/polypep-
tide coacervate phase or within coacervate microdroplets under
single turnover conditions (see Methods). Cleavage of the FRET-
substrate strand by HH-min increases the distance between 6-
carboxyfluorescein (FAM) and Black Hole quencher 1 (BHQ1),
resulting in increased fluorescence intensity. We further devel-
oped an inactive control ribozyme (HH-mut) by introducing two
point mutations at the catalytic site (see Methods).
HH-min (1 µM) and FRET substrate (0.5 µM) were incubated
within the CM-Dex : PLys bulk coacervate phase (4:1 final molar
concentration, pH 8.0). The RNA was then separated from the
coacervate phase and analysed by denaturing gel electrophoresis
(see Methods, Fig. 1). Excitingly, fluorescence gel imaging showed
the presence of cleavage product in the bulk coacervate phase
0
0.2
0.4
0.6
0.8
1
Cleaved fraction
02040
–0.05
0
0.04
0
0.2
0.4
0.6
0.8
1
0 200 400
Time (min)
0
0.2
0.4
0.6
0.8
1
Cleaved fraction
0 400
–0.02
0
0.01
0 200 400
Time (min)
0 200 400
Time (min)
0 200 400
Time (min)
0
0.2
0.4
0.6
0.8
1
C
A
C
A
C
C
G
U
U
3′
5′
A
A
A
G
G
U
A
G
G
A
C
C
C
U
G
G
A
G
G
G
A
G
U
C
U
G
A
G
A
C
U
A
U
C
C
G
G
G
A
G
C
U
FAM
BHQ1
(i) (iii)(ii)
HH-min
FRET-substrate
(i) (ii)
(i) (ii)
ab
c
d
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06072-w
2 NATURE COMMUNICATIONS | (2018) 9:3643 | DOI: 10.1038/s41467-018-06072-w | www.nature.com/naturecommunications
containing HH-min (Fig. 1b). In contrast, control experiments in
the absence of HH-min or in the presence of HH-mut showed no
evidence of the cleavage product, confirming that the wild-type
ribozyme drives substrate cleavage in the bulk coacervate phase
(Fig. 1b). The FRET assay was further exploited to characterise
the enzyme kinetics in both the bulk coacervate phase and buffer
by time-resolved fluorescence spectroscopy under single turnover
conditions by direct loading of HH-min and FRET substrate into
either cleavage buffer or bulk coacervate phase (see Methods).
The increase in fluorescence intensity of FAM was measured over
time and normalised to the amount of cleaved product generated.
Fitting the kinetic profiles of substrate cleavage in buffer
conditions with a single exponential revealed an apparent rate
constant, k
0
of 0.6 ± 0.1/min (Fig. 1c, N = 5), which was
comparable to the k
0
obtained in buffer analysed by gel
electrophoresis (0.38 ± 0.05/min) (see Methods, Supplementary
Fig. 2, N = 6) and to k
cat
values previously determined for a range
of hammerhead ribozymes (0.01–1.5/min)
34,35
. In comparison,
RNA cleavage within the bulk coacervate phase was clearly
biphasic (Supplementary Fig. 3A) with an observed faster rate
constant, k
1
, of 1.0 × 10
–2
± 0.1 × 10
–2
/min and a second slower
rate constant, k
2
, 1.5 × 10
–4
± 0.8 × 10
–4
/min (errors obtained
from 12 individual droplets). Thus, the fastest rate constant k
1
is
60-fold slower than in buffer conditions (k
0
= 0.6 ± 0.1/min)
indicative of reduced activity within the coacervate phase. In
addition, the transition to biphasic kinetics within the coacervate
phase compared to the aqueous buffer phase describes a different
kinetic mechanism of HH-min within the coacervate phase
(Fig. 1d). This may be attributable to heterogeneous ribozyme
populations with variations to secondary structure and/or
alternative conformational and equilibrium states, as observed
for some HH systems in aqueous buffer conditions
36,37
. Circular
dichroism (CD) spectra show a reduction in molar ellipticity ([θ])
for HH-mut within bulk coacervate phase compared to aqueous
buffer with a small commensurate shift in the peak maxima from
265 nm to 268 nm, respectively (Supplementary Fig. 4). These
results show that the fold of HH-mut is altered in the
polyelectrolyte-rich bulk coacervate phase, with an overall loss
of secondary structure that could affect catalytic activity. In
addition, it is possible that the charged and crowded coacervate
microenvironment restricts substrate binding, sterically hinders
substrate–enzyme complex formation and/or spatially restricts
diffusion of the cleavage assay components. Indeed, measured
diffusion coefficients of HH-min tagged with TAMRA (TAM-
HH-min) (1.0 ± 0.2 µm
2
/s) and FAM-substrate (1.6 ± 0.1 µm
2
/s)
in the bulk coacervate phase (Fig. 2, mean and standard
deviations are from at least two different samples with analysis
from at least 14 bleach spots from each experiment) from
Fluoresence Recovery after Photobleaching (FRAP) analysis
showed a significantly slower molecular diffusion of the ribozyme
and substrate compared to predicted diffusion coefficients of
RNA in buffer (~150 µm
2
/s, Fig. 2)
38,39
. The decreased mobility is
indicative of a more viscous and spatially restricted environment
in the interior of the coacervate phase (η = 114 ± 21 mPa. s,
Fig. 2c).
Hammerhead activity in coacervate microdroplets. To test the
activity of the ribozyme within individual droplets, the bulk
coacervate phase containing ribozyme and substrate was re-
dispersed in supernatant to produce microdroplets in solution
(see Methods). The final concentration of enzyme and substrate
in the microdroplet dispersion, formed from 1 µl of bulk coa-
cervate phase redispersed in 49 µl of supernatant was equivalent
to the final concentration of the bulk coacervate phase i.e. within
Fig. 1 Cleavage of the FRET substrate under different conditions. a HH-min
(black) and the FRET substrate (red). b Gel electrophoresis of RNA
cleavage in bulk coacervate phase (CM-Dex : PLys, 4:1 final molar ratio);
0.5 μM of FRET substrate was incubated with 1 μM of (i) HH-min, (ii) HH-
mut or (iii) no ribozyme in bulk phase (25 °C, 60 min). Samples were
analysed by denaturing PAGE followed by fluorescence imaging. The lack of
in-gel quenching of the FRET substrate likely results from modifications of
BHQ1 during PAGE
51
. c Real-time cleavage kinetics in 10 mM Tris-HCl pH
8.3 and 4 mM MgCl
2
. (i) A monoexponential fit (Methods, Eq. 3) (grey
line) to kinetic data (grey dots) and residuals of the fit (inset); (ii) mean of
the individual fits to each experiment (Blue line) with the standard deviation
of the mean of the fits (grey data points) (N = 5). d Cleavage in bulk
coacervate phase (normalised to the amount of cleaved product at t = 530
min from gel electrophoresis). (i) Biexponential fit (Methods, Eq. 4) (dark
grey line) to experimental data (grey dots) with the residuals (inset); (ii)
mean biexponential fit (orange) of individual fits (N ≥ 5). Grey data points
represent the standard deviation (N = 5) from the experimental data
0510
Time (s)
0.4
0.6
0.8
1
0510
–0.08
0
0.09
(i) FAM-substrate (ii) TAM-HH-min
–0.5 s
13 s
0 s
0510
Time (s)
0510
–0.05
0
0.07
(i) (ii)
0
0.5
1
1.5
0
50
100
150
(i) (ii)
I
norm
0.4
0.6
0.8
1
I
norm
D (μm
2
/s)
(mPa s)
a
b
c
Fig. 2 FRAP of bulk coacervate phase. Bulk coacervate phase (CM-Dex:PLys
(4:1 final molar ratio) containing either (i) 0.36 μM FAM-substrate or (ii)
0.36 μM TAM-HH-min. a Output frames from confocal imaging (63×) are
shown at t = −0.5 s before bleaching, directly after bleaching (magenta
circle, t = 0 s) and t = 13 s after bleaching. The fluorescence intensity was
normalised against a reference (green circle) and fit to standard equations.
Scale bars are 5 μM. b Plots of normalised FRAP data for HH-min (ii) and
FAM-substrate (ii) show the standard deviation (grey, N = 10) and fit
(blue) from the same bleach spot radius. c Diffusion coefficients and
viscosities obtained from b. Mean and standard deviations are from at least
two different samples with analysis from ≥14 bleach spots for each
experiment
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06072-w ARTICLE
NATURE COMMUNICATIONS | (2018) 9:3643 | DOI: 10.1038/s41467-018-06072-w | www.nature.com/naturecommunications 3
50 µl of phase under single turnover conditions (1 µM of HH-min
and 0.5 µM FRET substrate). Fluorescence optical microscopy
images showed an increase in FAM fluorescence intensity in the
droplets after 900 min (Fig. 3a).
Fitting the biphasic fluorescence signal allowed a direct
comparison of kinetic profiles between the coacervate micro-
droplet (Fig. 3b, Supplementary Fig. 3B) and bulk coacervate
phase environments. A modestly faster rate constant (k
1
and k
2
)
was observed in the microdroplets (k
1
of 4.4 × 10
–2
± 1.3 × 10
–2
/
min, k
2
of 2.3 × 10
–3
± 0.2 × 10
–3
/min, N = 12) compared to the
bulk coacervate phase (k
1
of 1.0 × 10
–2
± 0.1 × 10
–2
/min, k
2
of
1.5 × 10
–4
± 0.8 × 10
–4
/min, N = 11) (Supplementary Table 3,
Supplementary Fig. 5). Determination of the partition coefficients
of both the ribozyme and substrate (K
HH-min
= 9600 ± 5600
(N = 12) and K
HH-substrate
= 3000 ± 2000 (N = 20), Supplemen-
tary Fig. 6) by fluorescence spectroscopy (see Methods) showed
that both TAM-HH-min and FAM-substrate partition strongly
into the coacervate environment. 1 μl of bulk coacervate phase
was used to prepare coacervate microdroplet suspensions in a
total volume of 50 μl compared to 50 μl of bulk coacervate phase
for bulk experiments. Thus, based on the measured partition
coefficients, we calculated concentrations of 49.6 μM HH-min
and 24.3 μM substrate in a single microdroplet, compared to
1 μM HH-min and 0.5 μM substrate within the bulk coacervate
phase. Whilst the observed rate differences between the bulk
coacervate and coacervate microdroplet phases could be due to
variations in viscosity, quantitative FRAP analysis with two
different FAM-substrate concentrations (0.36 μM and 24.3 μM)
showed that the measured viscosities of the bulk coacervate phase
and coacervate microdroplet are comparable within error
(Supplementary Fig. 7). Therefore, secondary effects arising from
the increased RNA concentration within the coacervate micro-
droplet phase may be responsible for the increased rate constants
observed. The ribozyme and substrate concentrations are
approximately 50 times more in the coacervate microdroplet
(49.6 μM and 24.3 μM, respectively) compared to the bulk
coacervate phase (1 μM and 0.5 μM, respectively). The difference
in diffusion length scales of RNA in the microdroplet environ-
ment could lead to increased saturation of the ribozyme and
therefore greater apparent rate constants in the coacervate
microdroplets compared to bulk coacervate phase. In addition,
enrichment of RNA could lead to changes in material properties
such as water activity or dielectric constant, which have a direct
impact on the rate of hammerhead-catalysed RNA cleavage
40,41
.
Thus, secondary effects that result from the increased RNA
concentration within coacervate microdroplets may concomi-
tantly contribute to an increase in the apparent rate constant.
Selective RNA partitioning into coacervates. To further inves-
tigate the six-fold difference in the partition coefficient between
the ribozyme and substrate, we characterised the differences in
the rate and extent of sequestration of TAM-HH-min and FAM-
substrate from the surrounding aqueous phase into the droplet
after whole-droplet photobleaching. Coacervates containing
FAM-substrate (12-mer) showed complete fluorescence recovery
within 100 s and a recovery half time (τ) of 22 ± 3.5 s (N = 20). In
comparison, TAM-HH-min showed only 70% recovery after 300
s with τ = 189 ± 14 s (N = 11), attributed to either a low con-
centration of HH-min within the surrounding aqueous phase, a
slow rate of transfer into the coacervate droplet from its exterior
and/or immobilised ribozyme in the coacervate droplet, which is
unable to exchange with RNA in the surrounding aqueous phase
(Supplementary Fig. 8). The results from the FRAP experiments
complement the equilibrium partition coefficient. The 12-mer
substrate, with a lower partition coef ficient (K = 3000 ± 2000,
N = 20) shows a faster exchange between the droplet and
the surrounding aqueous phase compared with the 39-mer
ribozyme, which has a higher partition coefficient (K = 9600
± 5600 (N = 12)), shows slower exchange of RNA with the sur-
rounding environment. A strong correlation between FRAP half-
lives and partitioning coefficients was also described for RNAs in
other coacervate systems
4
.
To investigate additional sequence-dependent effects on
partitioning, we compared the partition coefficients of different
12-mer RNAs (Supplementary Fig. 6): FAM-substrate is pyr-
imidine rich but unstructured RNA; FAM-tet is a pyrimidine-rich
hairpin structure with a stable UUCG tetraloop; FAM-flex is an
unstructured purine-rich sequence (see Supplementary Table 2).
Our results show that for unstructured RNAs an increase in
purine content reduces partitioning of 12-mer RNAs (FAM-
substrate vs. FAM-flex) approximately 10-fold. Likewise, we
observe a decrease in the partition coefficient with an increase in
secondary structure for pyrimidine-rich RNA (FAM-substrate vs.
FAM-tet) (Supplementary Fig. 9). Thus, our results, and
others
4,42,43
, show that RNA sequestration and localisation
within coacervate droplets is dependent on the length, sequence
and also structure of the sequestered RNA.
For the RNAs specific to our HH ribozyme assay, the general
selective retention of longer length polynucleotides with transfer
of shorter length RNA can have interesting implications for
ribozyme catalysis within coacervate droplets. To investigate this,
we directly observed the localisation of TAM-HH-min and
FRET-substrate. CM-Dex:PLys coacervate micro-droplets (4:1
final molar ratio) containing TAM-HH-min were loaded into one
end of a capillary channel (Fig. 4a, region 1) while droplets
0 200 400 600 800
Time (min)
0
0.5
1
Cleaved fraction
Coacervate microdroplets
0 200 400 600 800
–0.05
0
0.05
0 5 10 15
0
10
20
30
Intensity (a.u.)
(i) (ii) (iii)
0 min 900 min0 min
Length (μm)
×10
3
0 5 10 15
0
10
20
30
Intensity (a.u.)
Length (μm)
×10
3
ab
Fig. 3 RNA catalysis in coacervate microdroplets. a (i) Wide-field optical microscopy images of CM-Dex:PLys (4:1 final molar ratio) coacervate
microdroplets prepared in cleavage buffer (1 μM of HH-min and 0.5 μM FRET substrate). Fluorescence microscopy images at t = 0 min (ii) and t = 900 min
(iii) show an increase in FAM fluorescence (see inset). Scale bars are 20 μM. b Background corrected and volume/endpoint normalised fluorescence
intensity of droplets. Standard deviation of kinetics from 12 micro-droplets (grey) with the mean biexponential fit (blue) and residuals (inset)
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06072-w
4 NATURE COMMUNICATIONS | (2018) 9:3643 | DOI: 10.1038/s41467-018-06072-w | www.nature.com/naturecommunications
containing the FRET-substrate were loaded into the other end of
the channel (Fig. 4a, region 3) in such a way as to prevent droplet
mixing whilst permitting passive diffusion of molecules through
the length of the channel (see Methods). Time-resolved
fluorescence optical microscopy images in both the TAM and
FAM channels were obtained at various locations along the
capillary channel (Fig. 4a, regions 1, 2, 3). Imaging over 500 min
in the TAM channel showed that, within the measurable
resolution, no diffusion of the ribozyme to droplets in other
regions of the channel occurs (Fig. 4b, region 2 and Supplemen-
tary Fig. 10). Conversely, over the time course, droplets in all
three regions exhibited increased FAM fl uorescence with droplets
in region 1 with the highest intensity and droplets in regions 2
and 3 with comparatively lower intensity (Fig. 4c). Analysis of
time-resolved images showed a delayed increase in the onset of
cleaved product fluorescence in region 2 and a further delayed
onset in cleaved product in region 3. These results are
commensurate with diffusion of the FRET-substrate out of the
droplets in region 3 and into droplets containing TAM-HH-min
in region 1 where cleavage takes place. The cleaved product then
diffuses out of the active droplets and into droplets in regions 2
and 3. Control experiments probing the transfer of FAM-
substrate only show increased fluorescence intensity in regions 2
and 3 giving a direct confirmation that the 12-mer substrate
diffuses between droplets (Supplementary Fig. 10). Taken
together, our results show that longer length RNA (HH-min) is
retained and spatially localised within the highly charged and
crowded interior of the coacervate droplet, while shorter RNAs
transfer between the droplets.
As other studies have shown that RNA rapidly exchanges from
PLys and adenosine triphosphate (ATP, Supplementary Fig. 1)
coacervate microdroplets into the surrounding environment
44
,we
also tested this coacervate system for selective localisation of
RNA. To this end, localisation experiments were undertaken as
previously described (see Methods) with PLys:ATP coacervate
microdroplets (4:1 final molar ratio) at pH 8: droplets containing
either TAM-HH min or FAM-substrate were loaded into one end
of a capillary channel, and coacervate droplets containing HH-
mut were loaded into the other end of the capillary channel in
such a way as to prevent droplet mixing (Supplementary Fig. 11).
Fluorescence optical microscopy images obtained in the middle of
the channel (Supplementary Fig. 11B, region 2) showed no
change in the fluorescence intensity of TAM-HH-min (Supple-
mentary Fig. 11C, i) over the course of the experiment (500 min).
In contrast, a small increase in the fluorescence intensity from
FAM-substrate (Supplementary Fig. 11C, ii) was observed after
300 min, suggesting a higher exchange rate of the 12-mer with the
environment. Both the partition coefficient and whole droplet
FRAP results for PLys:ATP coacervate microdroplets, containing
either TAM-HH-min (39-mer), FAM-substrate (12-mer) or
cleaved FAM-substrate (6-mer) (Supplementary Figs. 12, 13),
confirmed a consistent trend in RNA retention based on RNA
length with an order of magnitude difference in τ between the
different oligonucleotides (Supplementary Fig. 12). Whilst the
general trends are consistent with those observed with CM-Dex:
PLys coacervate microdroplets, a direct comparison of whole
droplet FRAP recovery times (Supplementary Table 4) between
the two microenvironments shows that RNA has a stronger
tendency to localise within the PLys:ATP microdroplets com-
pared to CM-Dex:PLys microdroplets. The longer recovery times
after whole droplet photobleaching in the PLys:ATP droplets
could be attributed to differences in molecular interactions of
RNA within the two microenvironments. For example, there may
be increased RNA–PLys interaction in the PLys:ATP environ-
ment compared to the PLys:CM-Dex, where presence of CM-Dex
could shield PLys–RNA interactions. In addition, favourable
Pi–Pi stacking interactions between the aromatic rings of ATP
and RNA would favour interactions of RNA within the PLys:ATP
system. Taken together, the results show that membrane-free
droplets prepared via coacervation offer general features such as
0 200 400
Time (min)
FAM channel
0
50
100
150
Intensity (a.u.)
Microscope channel
TAM-HH-min
FRET substrate
1
2
3
(i) Brightfield (ii) TAM channel (iii) FAM channel
1
2
3
1
2
3
ba
c
Fig. 4 Localisation and retention of RNA within coacervate droplets. a Schematic of localisation experiments where CM-Dex:PLys (4:1 final molar ratio)
coacervate droplets containing 0.36 μM(final concentration) TAM-HH-min were loaded into one end of a capillary channel (1). Droplets containing
0.36 μM FRET-substrate were loaded into the other end of the channel (3). b Wide-field optical microscopy images obtained using a 100 × oil immersion
lens in (i) bright field and fluorescence mode using filters for (ii) TAM or (iii) FAM. Images were captured in regions 1, 2 and 3 at t = 500 min (scale bar:
20 μM). c FAM fluorescence intensity. Shaded regions represent the standard deviation of at least seven droplets
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06072-w ARTICLE
NATURE COMMUNICATIONS | (2018) 9:3643 | DOI: 10.1038/s41467-018-06072-w | www.nature.com/naturecommunications 5
