High-throughput screening of filamentous fungi using nanoliter-range droplet-based microfluidics

Abstract

Filamentous fungi are an extremely important source of industrial enzymes because of their capacity to secrete large quantities of proteins. Currently, functional screening of fungi is associated with low throughput and high costs, which severely limits the discovery of novel enzymatic activities and better production strains. Here, we describe a nanoliter-range droplet-based microfluidic system specially adapted for the high-throughput sceening (HTS) of large filamentous fungi libraries for secreted enzyme activities. The platform allowed (i) compartmentalization of single spores in ~10 nl droplets, (ii) germination and mycelium growth and (iii) high-throughput sorting of fungi based on enzymatic activity. A 10⁴ clone UV-mutated library of Aspergillus niger was screened based on α-amylase activity in just 90 minutes. Active clones were enriched 196-fold after a single round of microfluidic HTS. The platform is a powerful tool for the development of new production strains with low cost, space and time footprint and should bring enormous benefit for improving the viability of biotechnological processes.

High-throughput screening of

OPEN

R A

P

Beneyton,* ,,* Postros , L Couvent M & Antoine

Aspergillus niger

Filamentous fungi are arguably the preferred source of industrial enzymes because of their excellent capacity for extracellular protein production1: lamentous fungi naturally secrete large amounts of hydrolytic enzymes, including amylases, cellulases or proteases, that are involved in the biochemical degradation of biomass. In addition, due to their extraordinary metabolic versatility, lamentous fungi are also used for the production of a variety of other products such as organic acids and pharmaceuticals2,3. Improving production strains is crucial for the economic viability of industrial biotechnology. Whilst genetic engineering and omics tools have made substantial contributions to rational strain improvement4, the lack of rapid, generic systems for screening of lamentous fungi still severely limits the discovery of better production strains and new enzymatic activites. Currently, the most widely used and exible screening methods are based on compartmentalizing clonal populations of lamentous fungi in microtiter plate wells. However, even using expensive automated colony pickers and liquid handling robots, throughput is limited to typically only ~100 fungi.h1 because of difficulties in manipulating lamentous fungi5. Fluorescence-activated cell sorting (FACS) can analyze and sort cells at rates of up to 7 104 cells.s1 6, but can only be used to sort lamentous fungi in the earlier stages of germination, owing to size incompatibility between lamentous fungi and the nozzle7. Futhermore, uorescence is detected in a continuous aqueous stream8 and the absence of compartmentalization makes it impossible to screen based on the activity of secreted proteins or extracellular metabolites.

Recently, droplet-based microuidics technology has allowed major advances for the screening of microorganisms by signicantly increasing the throughput and enlarging the range of systems that can be selected. Highly monodisperse droplets of picolitre volume can be made, fused, injected, split, incubated and sorted triggered on uorescence, at kHz frequencies911. Typically, single bacterial or yeast cells are compartmentalized in droplets of ~10 pl volume, allowing screening of enzymes expressed intracellularly12,13,14, on the surface of cells15 or secreted from cells16,17, with a 1,000-fold increase in speed and a 1-million-fold reduction in volume (and hence cost) compared to robotic microtiter plate-based systems15. Larger drops have also been used to grow yeast cells to screen for extracellular metabolite production or consumption (300 pL)18, or to screen hybridoma cells for monoclonal antibody production (660 pL)19, albeit at lower throughput (110 s1). However, microuidic HTS have never

2 *

SCIENTIFIC REPORTS

1

www.nature.com/scientificreports/

Figure 1. Growth of lamentous fungi in water in oil (w/o) droplets. (a) Microscopic image sequence (40) over 17h of a 250 pl droplet containing a single germinating spore. The droplet was immobilized in a drop spot chip45 and incubated at 30C. (b) Micrograph (15) of 18nl droplets aer single spore encapsulation and 24h incubation at 30C. (c) Micrograph (15) of 18nl droplets aer single spore encapsulation and 32h incubation at 30C. (d) Epiuorescence image (10) of 18nl droplets aer single spore encapsulation with a uorogenic substrate and 24h incubation at 30C: the secreted -amylase activity induces green uorescence within droplets containing fungi. Red circles highlight the hyphal tips exiting the droplet.

been used with lamentous fungi. Indeed, growing and screening lamentous fungi in droplets is challenging due to the apical growth of the hyphae aer spore germination to form an expanding branched mycelia network20. As we show here, lamentous fungi cannot be screened in picoliter droplets in microuidic systems: the hyphal tips, where most protein secretion occurs21, rapidly exit the droplets, leading to uncontrolled coalescence. This limits the incubation period to only a few hours, which is too short to allow secretion of enough enzyme for screening.

This work proposes a novel approach for the HTS of lamentous fungi using nanoliter-range droplet-based microuidics tools. Single spores can be encapsulated in ~10nl droplets (volume ~3 orders of magnitude larger than those typically used for sorting bacteria and yeast) which can be incubated and sorted based on uorescence. This system allowed 24h incubation of lamentous fungi (Aspergillus niger) and screening of ~7,000 fungi. h1 based on the activity of a secreted enzyme (-amylase). A whole-genome UV-mutated A. niger library was enriched 196-fold for fungi secreting -amylase aer one round of microuidic HTS. Sorted clones were further analyzed using a robotic microplate based system at ~400 clone.h1: 98.1% exhibited -amylase activity. The HTS platform can easily be adapted to screen for other enzymatic activities.

Using nanoliter size droplets is essential for screening of lamentous fungi. Typically, in asks, the time from which Aspergillus spores germinate to the stage where they start to display detectable enzymatic activity is ~24 h and optimum levels of activity are usually only reached aer several days of incubation, depending on the secreted enzyme. The requirement for long incubation time combined with the rapid growth of the fungal hyphae makes screening lamentous fungi in picoliter droplets a challenging task: even in 250 pl droplets the hyphal tips exit the droplets in ~15h, causing uncontrolled coalescence: (Fig.1a, Supplementary Movie S1). In contrast, encapsulating single spores in 18nl droplets allows growth of the branched mycelial network for up to 24h conned in the droplet (Fig.1b) with the hyphal tips exiting the droplets only aer incubation for 32h (Fig.1c). Furthermore, aer 24h incubation, secreted -amylase was easily detectable in the droplets using a uorogenic substrate (Fig.1d).

Increasing droplet volume from the pl to the nl range aects droplet formation, stability and dielectrophoretic sorting. It was straightforward to produce highly monodisperse (polydispersity 5% by volume) 1020 nl droplets at 8090 droplets.s1 using a ow-focusing22 device with a constriction orice 225m wide and 250m deep (Supplementary Fig. S1 and Movie S2), an oil phase of Novec HFE-7500 uorinated oil containing 2.5% (w/w) KryJeD900 (a triblock copolymer uorosurfactant)23 and an aqueous phase of synthetic growth medium.

Droplets were found to be highly stable and could be collected, incubated and reloaded. Analysis of the size distribution of 10 nl droplets aer 24 h incubation at 30 C in the form of a concentrated, creamed emulsion in the glass capillary and re-injection into the sorting chip (Supplementary Fig. S1) indicated that 1.0% of droplets were coalesced (Supplementary Fig. S2).

We developed a uorescence-activated droplet sorting (FADS) device specically adapted to sort nanoliter droplets. In dielectrophoretic droplet sorters12,15,24, droplets ow in carrier oil towards a Y-shaped junction. With no electric eld, all drops ow into the waste channel which oers lower hydrodynamic resistance than the second, collect channel. To direct droplets into the collect channel, on chip electrodes are energized, creating an electrical eld gradient, which generates a dielectrophoretic force (DEP) acting on the droplets. In order to be

SCIENTIFIC REPORTS

2

www.nature.com/scientificreports/

sorted, DEP forces must displace the droplet by a critical distance orthogonal to the ow, do, in the time, tp, it takes

the droplet to traverse the electrical eld gradient. The DEP force on a spherical droplet is given by equation (1)25,

[combarrowextender][combarrowextender]

F R Re f E

4 [ ] (1)

DEP m CM

in which, |FDEP| is the magnitude of the DEP force, R is the radius of the droplet being subjected to the DEP force, Re [fCM] is the Claussius-Mosscoti factor (which itself depends on ip and im: the complex permittivity of the droplet and the carrier uid, respectively) and [combarrowextender][arrowrightnosp]

E 2 denotes the electrical eld gradient. Assuming that d0=2R, and that, in order to prevent false positives, no more than one droplet can be in the electric eld in the sorting chamber at any one time, the theoretical maximum sorting frequency, fsort, can be calculated (see Supplementary

Information):

m CM

where designates the viscosity of the carrier oil. Equation(2) shows that fsort is directly proportional to R and to [combarrowextender][arrowrightnosp]

E 2, and that increasing R apparently favors DEP sorting as lower [combarrowextender][arrowrightnosp]

E 2 is required. However, larger droplets are more easily split by electric elds26. The electric eld, Emax, above which droplets split, when Maxwell stress surmounts the resistance to deformation due to interfacial tension, is given by the expression:

E = R (3)

max

[combarrowextender][arrowrightnosp][notdef]= [combarrowextender][arrowrightnosp]

[tildenosp2char]

3 2

f R Re f E

sort

2

=

[combarrowextender][arrowrightnosp]

[ ]

3 (2)

0

Therefore, the maximum sorting frequency is a function of both the magnitude, E, and inhomogeneity, [combarrowextender][arrowrightnosp]

E 2,

of the electric eld in the sorter and depends on the volume of the drops and the size and geometry of the sorter.

We investigated two dierent electrode congurations: the rst design had both charged and ground electrodes on the same side of the sorter (Fig.2a), as used for uorescence-activated sorting of pl droplets12,15 and

the second design had the charged and ground electrodes on the opposite sides of the sorter (Fig.2b). Finite element analysis (FEA) simulations indicated that in the same-side electrode conguration, although the DEP force was high between the positive and ground electrodes close to the channel wall adjacent to the electrodes, it drops o rapidly across the width of the channel (Fig.2a and Supplementary Fig. S3). In contrast, using the cross-side electrode conguration the DEP force extends across the entire width of the sorting channel (Fig.2b and Supplementary Fig. S3). By combining a model of the electric eld distribution with equations(2) and (3) it was possible to calculate the maximum droplet sorting frequency limited by electrosplitting, fe, as a function of droplet volume (Fig.2c; Supplementary Information). With both electrode congurations fe is inversely proportional to the droplet volume. However, if the droplet volume is >158 pl (~67 m diameter), the cross-side electrode conguration allows the highest sorting rate, whereas with droplets of <158pl the same-side electrode conguration allows the highest sorting rate.

The superiority of the cross-side electrode conguration for sorting 10 nl droplets was conrmed experimentally. Sorting of 10 nl droplets was achieved when AC electric eld pulses of 1.4 kVpp, 30 kHz, 30 ms, were applied with the cross-side electrode conguration (Supplementary Movie S3). In contrast, the same electric eld pulses could not sort droplets with the same-side electrode conguration (Supplementary Movie S4). The efficiency of the sorting device with the cross-side electrode conguration was further investigated by sorting 10nl droplets of a binary emulsion comprising low red uorescent droplets containing 1M sulforhodamine B (negative; 65%) and highly red uorescent droplets containing 30M sulforhodamine B and blue ink (positive; 35%) (Fig.2d). The false positive and false negative error rates were determined either by color imaging of the emulsion before and aer sorting or by high speed video analysis of droplets trajectories during the sorting process at dierent throughputs. When droplets were sorted at 4s1, the emulsion collected in the positive channel was composed of 100% positive (blue) droplets while the emulsion collected in the negative channel contained 99.9% of negative droplets (Fig.2e; Supplementary Fig. S3). Based on the video analysis of the sorting process (Supplementary Movie S5), the false positive and false negative error rates were found to less than 0.07% and equal to 0.07%, respectively (Fig.2f). These low error rates validate the ability of the cross-side electrode conguration device to efficiently sort 10 nl droplets based on uorescence. The few false negative errors were always due to instability in the reloading regime of the droplets, resulting in temporary modication of the spacing at the sorting junction. The device could operate at up to 21 droplets.s1 with acceptable false positive and false negative error rates (Fig.2f; Supplementary Movie S6), and was also able to sort 20nl droplets with similar efficiency (Supplementary Fig. S3). This is close to the maximum theoretical sorting frequency, fe, for 10nl droplets calculated from the model (Fig.2c), which was 3 droplet.s1 for same-side electrode conguration and 46 droplet.s1 for the cross-side electrode conguration.

The screening platform for Aspergillus niger (Fig.3a) was composed of two distinct microuidic devices (Supplementary Fig. S1). A drop maker allowed the encapsulation of single spores with a uorogenic -amylase substrate (Supplementary Movie S2). Droplets of 18 nl were produced at ~80Hz by hydrodynamic ow focusing of a spore suspension with a uorinated oil phase containing 2.5% (w/w) KryJeD900 uorosurfactant. Fluorinated oils have been shown to facilitate respiratory gas delivery to both prokaryotic and eukaryotic cells in culture27. The number of spores per droplet follows a Poisson distribution28 and was controlled by adjusting the initial density of the spore suspension to give an average number of spores per drop , of 0.10.3. The droplets were collected in a glass capillary and incubated for

c

SCIENTIFIC REPORTS

3

www.nature.com/scientificreports/

Figure 2. Validation of the sorting device. (a) Same-side electrode conguration: schematic and DEP force simulation (b) Cross-side electrode conguration: schematic and DEP force simulation. The corresponding electrical potential distributions for each electrode conguration are shown in Figure S2. (c) Calculated maximum sorting frequency fe as a function of droplet volume for same-side and cross-side congurations. (d) Fluorescence histogram of a binary emulsion of 10nl droplets containing 1M sulforhodamine B (population 1, 65%) or 30M sulforhodamine B plus blue ink (population 2, 35%) before sorting using the DEP sorter with the cross-side electrode conguration. The sorting gate was set to sort only droplets with red uorescence between 40 and 55 RFU (vertical dashed lines) corresponding to population 2. (e) Color pictures of the emulsion before (i) and aer sorting at 4 droplet.s1 from the negative (ii) or positive (iii) channels. (f) Video analysis of the sorting efficiency at dierent throughputs.

24 h at a 30 C to allow germination, hyphal growth and -amylase secretion within the droplets. Aer incubation, the droplets were re-injected into a sorting device with cross-side conguration electrodes (Supplementary Fig. S1 and Movie S7) which was used to sort droplets based on -amylase activity (Supplementary Movie S8). The uorogenic substrate was starch labeled with multiple BODIPYFL uorophores which are auto-quenched

until the starch is hydrolyzed, resulting in an increase in green uorescence (Fig.3b). Sorted fungi were recovered, sporulated and further characterized. The microuidic platform was applied to screen for Aspergillus niger strains which overproduce -amylase from large whole-genome mutated libraries. The Aspergillus niger strain O58, a Soufflet -amylase producing reference strain was exposed either to a chemical agent (2% N-Methyl-N

SCIENTIFIC REPORTS

4

www.nature.com/scientificreports/

Figure 3. Droplet-based microuidics screening platform. (a) Schematic of the system. 1. Generation of a whole-genome mutated fungi library. 2. The spores are compartmentalized in ~10nl droplets with a uorogenic enzyme substrate using an appropriate spore to droplet ratio allowing compartmentalization of single spores. 3. The emulsion is incubated 24h o-chip at 30C in a glass capillary to allow fungi development, enzyme secretion and substrate digestion within the droplets. 4. Droplets are reloaded into a sorting device and sorted based on enzymatic activity according to uorescence intensity. 5. Fungi are recovered from sorted droplets and either characterized or subjected to another round of mutagenesis and/or selection. (b) -amylase uorogenic assay. The substrate consists of a starch backbone with multiple quenched BODIPYFL uorophores. -amylase

hydrolyzes the starch backbone to unquench the uorophores and induce uorescence.

-Nitro-N-Nitrosoguanidine [MNNG]) or to UV light (254 nm, 1.30 mW.cm2) to introduce mutations in the entire genome. The diversity of chemical and UV libraries was 9.106 and 5.104, respectively. The -amylase activity of the O58 strain, the chemical library (2 and 3 h exposure) and the UV library (30, 60 and 90 s exposure) were analysed with the microuidic platform (Fig.4a and Supplementary Fig. S4). The phenotypic signature of the O58 strain presents an active population [6.2% of drops (theoretically 9.5% with = 0.1); 2.5 17.6% RFU] well separated from empty drops or drops showing no -amylase activity [93.8% of drops; 1 7% RFU] (Fig.4a). Exposure to MNNG for 2h resulted in an only slightly modied phenotypic signature [23% of active drops (theoretically 19.7% with =0.22); 2.42 15.7% RFU] (Fig.4a). In contrast, the phenotypic signature was heavily impacted by 60 s UV-exposure, and the library predominantly comprised variants exhibiting no or low -amylase activity and only a small number of variants with phenotypes giving equal or higher green uorescence (-amylase activity) than the O58 strain [2.6% of active drops (theoretically 19.7% with =0.22); 1.9350.3% RFU] (Fig.4a).

Based on these results, the UV mutagenesis method (60 s exposure) was chosen for the subsequent screening steps. The high mutation rates with UV mutagenesis result in high mortality (60s UV exposure results in a survival rate of 2.6% (Supplementary Fig. S4)) and a large number of inactive clones. However, a large fraction of inactive clones is also statistically associated, in a random whole-genome mutagenesis strategy, to a higher probability of targeting the desired pathway or gene.

A library was constructed using UV-mutagenesis (60s exposure) of the O58 mother strain. The library was encapsulated in 18nl droplets with the uorogenic -amylase substrate. The number of viable spores per droplets (21% occupancy; = 0.24) was controlled by adjusting the spore density based on the survival curve (Supplementary Fig. S4). Aer 24h incubation at 30C, droplets were sorted at 10Hz based on uorescence, selecting droplets containing -amylase producing fungi (Fig.4b; green region) (Movie S8). Around 5.104 droplets (~104 fungi) were screened in about 90min, 1.45% were sorted (~750 droplets) and 616 fungi were recovered on an agar plate. The enrichment of the screening steps was evaluated by analyzing sorted fungi at the single clone level (~7,000 fungi) (Fig.4c). The activity distributions show an important (196-fold) enrichment for active fungi: more than 97% of sorted fungi showed -amylase activity, the enrichment being limited by co-encapsulation events during the emulsication step12.

The sorted fungi population was screened using a microtitre plate-based robotic platform on an industrial relevant solid medium based on rapeseed meal. 1296 colonies were picked into 24-well plates, grown for 5 days and secreted -amylase activity was measured in 96-well plates using the BODIPYFL based substrate (Fig.5a and Supplementary Fig. S5). 144 replicates of the

O58 fungi were screened as positive controls. In total, 1,272 sorted fungi were analysed (87.2% of the sorted population): 98.9% were founded to be active while only 1.9% did not show detectable -amylase activity. Most of the

SCIENTIFIC REPORTS

5

www.nature.com/scientificreports/

Figure 4. Screening Aspergillus niger libraries for -amylase secretion. (a) Phenotype distribution (-amylase activity) within large fungi populations. Histograms of the green uorescence signal (514nm; corresponding to -amylase activity) of the droplets aer incubation at 30C for 24h of the O58 strain (i),a chemically mutated library (2h exposure) (ii) or a UV-mutated library (60s exposure) (iii). The wild-type histogram is displaying ~200,000 droplets (~20,000 fungi). The library histograms are each displaying ~30,000 droplets (~6,000 fungi). (b) Histogram of the green uorescence signal (514nm; related to -amylase activity) of the droplets before sorting. To maximize enrichment for active variants, droplets were sorted (green region) only if green uorescence was at least twice the modal uorescence of the main population (empty droplets) i.e. >1.9 RFU. (c) Histogram of the green uorescence signal (514nm; related to -amylase activity) of the selected fungi population. For all histograms, if is the standard deviation of the green uorescence of the main population (empty droplets), droplets are considered (+) if RFU >3 and () if RFU 3 (red line threshold). The number of droplets is converted into the number of strains based on the occupancy.

sorted fungi showed an activity level similar to O58 (47.3%) while 37.3% showed a higher activity (>1O58). We measured the skewness of each distribution as the Pearson coefficient of skewness. We obtained a value of 0.5 for the O58 distribution, meaning that the distribution is not signicantly skewed compared to a normal distribution (p=0.0124; = 0.01), and a value of 1.05 for the sorted fungi distribution, indicating that the distribution has a signicant positive skew (p<0.00001; = 0.01) (Supplementary Information).

We analysed in replicate 10 strains from the plate screen with -amylase activity 3 by larger scale shake-ask fermentation. This format is closer to industrially relevant conditions and normally precedes further evaluation and scale-up in fermenters. The -amylase activity was measured aer 6 days of fermentation in PGS medium (100 ml) and compared to the O58 strain activity (Fig.5b). All the strains tested except one displayed activity similar to or higher than O58, the best (M8) being 2.3-fold higher than the parental strain. The activities measured did not correlate exactly with those from microtitre plates, however, probably due the dierent production conditions (solid-state fermentation on rapeseed meal versus shake-ask fermentation in liquid synthetic PGS medium).

We have developped and applied an efficient new approach for the high-throughput screening of filamentous fungi based on secreted enzyme activities. Using droplet-based microuidic tools adapted to manipulate nanoliter-volume droplets, the technique allows fast enrichment of large population of fungi and can be integrated to industrial screening processes to speed up the discovery of new production strains of biotechnological interest.

Compared to automated colony pickers and liquid handling systems, the microuidic platform oers higher throughputs, lower reagent consumption and reduced spatial footprint, all of which contribute to lower screening costs (Supplementary Table S1): screening 104 A. niger variants directly using the robotic microtitre plate-based screening platform would have taken more than 16 days and cost $8,770, whereas screening 104 variants using the microuidic system took less than 24h and cost $14 (and the subsequent screening of 1,272 sorted variants using the microtitre plate-based system took only 8 days and cost $1,100), costs being estimated based on consumables only. In addition to inherent throughput and scale advantages related to the microuidics format, this approach also greatly simplies the manipulation of lamentous fungi: o-chip fungi handling is limited to the manipulation of spores and operations on mycelia are performed by manipulating droplets.

Combining microuidic and robotic microtitre-plate HTS is an attractive strategy. Here, single fungal spores were compartmentalized in droplets together with a uorogenic -amylase substrate, and the droplets and the germinated fungi they contain were sorted aer 24h incubation, triggered on uorescence. However, with such

SCIENTIFIC REPORTS

6

www.nature.com/scientificreports/

Figure 5. Microtiter-plate and shake-ask fermentation assays. (a) Relative -amylase activity of the O58 strain (144 replicates) and the strains selected using the microuidic screening (1,272 clones), analyzed using a robotic microtitre plate screening system. The standard deviation of the O58 strain is indicated. (b) Histogram displaying the -amylase secretion level of 10 of the most active fungi from the plate screen and the wild type O58 strain aer 6 days of cultivation at 30C in 100ml of PGS medium. Mutants (gray) and O58 (red) were ranked from the lowest to the highest -amylase activity. All fermentations were performed in duplicate and error bars correspond to 1 standard deviation.

an endpoint measurement, it is difficult to distinguish reliably between variants exhibiting dierent levels of -amylase activity as fungal growth, enzyme secretion and the uorogenic -amylase reaction are coupled, and in many droplets all (or most) of the substrate was consumed. However, sorting all variants with a uorescent signal equal to or greater than the mother strain using the microuidic system efficiently removed variants showing reduced or no -amylase activity: before sorting, 83% of clones showed little or no -amylase activity, but a single round of microuidic sorting resulted in a 196-fold enrichment of active clones. The throughput of microuidic HTS enables the screening of highly mutated libraries, which are mainly composed of inactive clones, but where there is also a higher probability of nding benecial mutations, as the higher mutation rate increases the chance of hitting the target gene(s).

Furthermore, the microuidic assay could be made more quantitative by uncoupling fungal germination, growth and enzyme secretion from the enzymatic assay. This could be achieved by adding the uorogenic substrate to droplets aer incubation to allow fungal growth and enzyme secretion using droplet electrocoalescence29

or picoinjection30. It may also be possible to increase the throughput above ~7,000 fungi.h1. The maximum experimental sorting rate (21Hz) is close to the calculated theoretical maximum sorting rate limited by DEP and electrocoalescence (46Hz). However, the sorting frequency was not limited by DEP forces or electrosplitting, but by purely uidic constraints: higher reinjection frequencies resulted in droplets breaking when oil was added to space the reinjected droplets. Higher sorting rates could potentially be achieved by reducing the oil spacer ow rate, a strategy which, combined with other sorter design improvements, allowed sorting of pl volume droplets at 30kHz31. Throughput could be further increased, at least to a certain extent, by increasing (at the expense of increasing the number of false positives due to co-compartmentalization of more than one variant in the same droplet12).

The microuidic system is not limited to screening for -amylase activity: lamentous fungi can potentially be screened for the production of other enzymes or metabolites. The only restriction is that there must be a uorogenic assay for the desired enzymatic activity or metabolite, and the metabolite and uorescent product must not exchange between droplets over the time of the experiment. The use of uorinated oils for the continuous phase mitigates against exchange, since non-uorinated molecules are highly insoluble in uorinated oils32. However, exchange can also be mediated by micellar transport33. Nevertheless, many microtitre-plate-based assays can be transposed directly to droplets, as was the case with the -amylase assay used here, and direct, or coupled enzymatic assays can be used. Many other uorogenic assays can also be simply adapted for microuidic systems by chemically modifying the uorogenic substrate to increase the hydrophilicity of the uorescent product34,

which reduces partitioning into micelles in the continuous phase and exchange between droplets33. For example,

SCIENTIFIC REPORTS

7

www.nature.com/scientificreports/

droplet-based microuidic assays have been developed to screen for bacteria and yeast producing cellulases, based on either directly screening for exogluconase (cellobiohydrolase) activity using a uorogenic assay35 or

using a coupled uorogenic enzyme assays to measure endogluconase activity16, and have been used for the ultrahigh-thoughput bioprospection of natural cellulolytic bacteria35. In the future, novel assays based on, for example, changes in osmotic pressure36,37 or mass spectroscopy3840 might also be developed.

The ability to screen large lamentous fungi populations in a few hours at low cost should bring enormous benet for identication of new biotechnologically interesting lamentous fungi. Besides this promising biotechnological application, this screening platform would also be applicable to more fundamental studies in combination with whole genome sequencing to help identify mutations and thereby regulatory networks responsible for improved phenotypes. The understanding of those genetic factors might be used to optimize lamentous fungi via inverse metabolic engineering41.

Poly-(dimethylsiloxane) (PDMS) microuidic devices were fabricated as previously described42 from 250m-deep molds of SU-8 2150 negative photoresist (MicroChem Corp).

The optical setup used to monitor microuidics experiments was previously described, as well as the data acquisition and control system34.

In addition, to allow the sorting of a particular droplet, the data acquisition card provided a signal to a model 623B high-voltage amplier (Trek Inc.) connected to the electrodes of the microuidic device.

We used aqueous droplets in Novec7500 uorinated oil (3M) stabilized against coalescence by a triblock copolymer uorosurfactant comprising two peruoropolyether (PFPE) chains linked by one Jeamine polyetheramine chain (PEA), KryJeD900. KryJeD900 surfactant was prepared in house from

the commercially available carboxylic acid Krytox157-FSH (Dupont) and Jeamine polyetheramine (ED 900,

Huntsmann) based on the synthesis route described in23,43. Briey, Krytox157-FSH (50 g; 7.8 mmol assuming 6500g.mol1) was dissolved in 150ml of CaCl2 dried Novec7100 oil under a N2 atmosphere. Next, oxalyl chloride (7.7ml; 90mmol) was added dropwise and the mixture stirred overnight at 70 C. Aer evaporating the solvent, the resulting product was dissolved in 100ml of FC-3283 oil (3M). Jeamine polyetheramine (3.3ml; 7.9 mmol),

20ml of dried tetrahydrofuran, thriethylamine (1.6ml; 11.5 mmol) and 40ml of dried tetrahydrofuran were subsequently added in a twin-neck round-bottom ask under a N2 atmosphere. The Krytox157-FSH acid chloride solution was added in the ask, followed by 20 ml of FC-3283 oil (3M). The mixture was stirred overnight at room temperature. Aer removing the solvents and purication by ltration, the product was used directly in the experiments.

Liquids were pumped into the microuidic devices using standard-pressure infusion-only PHD 22/2000 syringe pumps (Harvard Apparatus Inc.). Syringes (Omnix-F; BBRAUN) were

connected to the microuidic devices using 1.2 40 mm needle (Terumo) and PTFE tubing (Fisher Scientic) with an internal diameter (ID) of 1.06mm and an external diameter (OD) of 1.68mm. Droplets were produced using a dropmaker device (Supplementary Fig. S1) by ow-focusing of the aqueous stream with two streams of Novec7500 oil containing 2.5% (w/w) of KryJeD900 surfactant. The device was used to produce 10 nl droplets (Qaqueous 3 ml.h1, Qoil 10 ml.h1, 80 droplets.s1), 18 nl (Qaqueous 6 ml.h1, Qoil 8 ml.h1, 90 droplets.s1) or 20 nl droplets (Qaqueous 4.6 ml.h1, Qoil 5 ml.h1, 90 droplets.s1) depending on the experiment. The generated emul

sions owed o-chip through PTFE tubing to a glass capillary attached to a Peltier device for o-chip incubation44. Droplets were reloaded (Qdroplets 125700l.h1) in a uorescence activated droplet sorting (FADS) device (Supplementary Fig. S1) and spaced with Novec7500 oil (Qoil 1021 ml.h1). The droplets were analysed by the optical setup and uorescent droplets were sorted at 4 to 20 droplet.s1 by applying AC eld pulses (30kHz; 1400 1800 Vpp; 1040 ms). The two collection outlets were connected to TYGON tubing (7.94mm OD; 4.76 mm ID)

via 6mm-diameter L-shaped connectors (VWR).

A binary emulsion comprising 10 nl droplets was produced by mixing two emulsions produced successively using the dropmaker device. The rst droplet population contained 1 M of sulforhodamine B in YGC medium (Yeast extract 5 g.l1, Glucose 20 g.l1, Chloramphenicol 0.1g.l1) and the second contained 30M of sulforhodamine B and blue ink colorant in YGC medium. The binary emulsion was reloaded into the sorting device and sorted as a function of red uorescence to select only those droplets containing 30 M of sulforhodamine B. Both sorted and unsorted droplets were collected. The performance of the sorter was evaluated before and aer sorting either by imaging the emulsion on a glass slide using a color camera (Nikon DS-Fi2) or by video analysis of the sorting process using a high-speed camera (Mikroton Eosens MC1310).

The Aspergillus niger O58 strain was obtained from the Ets J. Soufflet strain collection. The O58 strain was plated on a PDA (AES Laboratoire) plate and grown at 30C for 10 to 15 days for optimal sporulation. Spores were suspended in 20ml of PGS medium (Glucose 10g.l1,

Pancreatic peptone 6g.l1, MgSO4.7H2O 0.5g.l1, KH2PO4 0.5g.l1, FeSO4.7H2O 0.5mg.l1). The spore suspension was stirred for 10 min, ltered through sterilized gauze to remove mycelium residues and stirred again for 10min. The spore density was estimated using a Thoma cell counting chamber.

Aer sporulation, spores were suspended in 0.01% Triton X-100 (20ml). The spore suspension was stirred for 10 min, ltered through sterilized gauze and stirred again for 10 min with glass beads to

SCIENTIFIC REPORTS

8

www.nature.com/scientificreports/

homogenize. The spore concentration was adjusted to 3.106 spores per ml. 3ml of the spore suspension was spread on a glass petri dish and exposed to 6 W UV light (Dymax PC-2000/38003) at 254 nm for 60s at a distance of 5cm from the lamp (1.30mW.cm2) to introduce mutations randomly throughout the whole fungi genome. The procedure was repeated three times to UV treat a total of 9ml of spore suspension. The UV-treated spore suspension was centrifuged 10min at 2500g and the supernatant was discarded. The pellet was suspended in PGS medium.

As for UV mutation, a spore suspension was prepared. The spore concentration was adjusted to 1.107 spores per ml. 5 ml of 0.2 M phosphate buer (pH 6.2) containing 6% (w/w) of N-Methyl-N -Nitro-N-Nitrosoguanidine (MNNG) was added to 10 ml of spore suspension. The suspension was stirred at 30C for the desired mutagenesis time. Mutated spores were then centrifuged 5min at 4500g and the pellet was suspended in 5ml of Triton X-100 (1g.l1). The procedure was repeated three times to wash away the mutagenesis agent. The pellet was nally suspended in PGS medium.

The spore suspension was diluted to the appropriate spore to droplet ratio to obtain typically 0.10.2 spore per droplet in PGS medium containing the BODIPYFL-labeled DQTM starch substrate (50g.ml1; Enzchek Ultra Amylase Assay Kit; Life

Technologies) and Sulforhodamine 101 (5M). Spores were encapsulated in 18nl droplets using the dropmaker device. The emulsion was collected in a glass capillary, incubated o-chip for 24h at 30C and then reloaded in a FADS device for uorescence analysis.

The library spore suspension was encapsulated in 18nl droplets as described in the previous paragraph. The emulsion was collected in a glass capillary, incubated o-chip for 24h at 30C and then reloaded in a FADS device for uorescence analysis and selection of droplets containing active fungus at 10 droplets.s1 by applying AC eld pulses (30kHz; 1600Vpp; 35ms).

Sorted fungi were spread on PDA plates, grown for 48 h at 30 C, counted, and then grown for an additional 6 days to allow sporulation. For conservation, spores were suspended in 20ml of water, counted on a Thoma cell counting chamber and frozen at 80C in 15% glycerol at a density of at least 106 spores per ml.

The sorted fungi population and O58 strain were grown on agar plates to feed a Hamilton Microlab STAR 8/96 ML picking robot. Spores were plated on PDA plates (30 plates for the O58 strain and 150 plates for the sorted population) and grown for 40h at 30C to give 15 to 20 thalli per plate. The picking robot was run for 3 days to transfer every strain into 24-wells plates containing agar growth medium based on rapeseed meal (rapeseed meal 100g.l1; pancreatic peptone 6g.l1; MgSO4 7 H2O 0.5g.l1; KH2PO4 0.5g.

l1; FeSO4 7H2O 0.5mg.l1, Agar 20g.l1). Plates were incubated for 5 days at 30C, 95% hygrometry. 1ml of water was then added to each well and plates were incubated 48h at 4C for protein extraction. Aqueous extracts were transferred in 24-well plates for automated titration of -amylase activity using a Hamilton Microlab STARlet 4 ML liquid handling robot. The uorescence read out was performed in 96-wells plates over 10min at 45C from 25l of sample added to 25l of BODIPYFL-labeled DQTM starch substrate (50g.ml1; Enzchek Ultra Amylase

Assay Kit; Life Technologies; ex 480nm; em 515nm) in citrate buer (0.33M, pH3.3). A calibration curve was produced using an -amylase standard (7500U.g1) for the 0300mU.ml1 range (Supplementary Fig. S5).

For each strain, two 2.5l asks containing 100ml of PGS medium were each inoculated with 1ml of a glycerol stock spore suspension (5.108 spores.ml1). The asks were incubated at 30C (150 rpm) for 6 days. Supernatants were titrated for -amylase activity. The uorescence readout was performed in 96-wells plates over 10 min at 45C from 25l of sample added to 25 l of BODIPY

FL-labeled DQTM starch substrate (50g.ml1; Enzchek Ultra Amylase Assay Kit; Life Technologies; ex 480nm; em 515nm) in citrate buer (0.33 M, pH3.3).

References

1. Punt, P. J. et al. Filamentous fungi as cell factories for heterologous protein production. Trends Biotechnol. 20, 200206 (2002).2. Lubertozzi, D. & Keasling, J. D. Developing Aspergillus as a host for heterologous expression. Biotechnol. Adv. 27, 5375 (2009).3. Meyer, V. Genetic engineering of lamentous fungi - Progress, obstacles and future trends. Biotechnol. Adv. 26, 177185 (2008).4. Meyer, V., Wu, B. & Ram, A. Aspergillus as a multi-purpose cell factory: current status and perspectives. Biotechnol. Lett. 33, 469476 (2011).

5. Alberto, F., Navarro, D., De Vries, R. P., Asther, M. & Record, E. Technical advance in fungal biotechnology: Development of a miniaturized culture method and an automated high-throughput screening. Lett. Appl. Microbiol. 49, 278282 (2009).

6. Eisenstein, M. Cell sorting: Divide and conquer. Nature 441, 11791185 (2006).7. Throndset, W. et al. Flow cytometric sorting of the lamentous fungus Trichoderma reesei for improved strains. Enzyme Microb. Tech. 47, 335341 (2010).

8. Shapiro, H. Practical Flow Cytometry 4th edn (Wiley-Liss, 2003).9. Guo, M. T., Rotem, A., Heyman, J. A. & Weitz, D. A. Droplet microuidics for high-throughput biological assays. Lab Chip 12, 21462155 (2012).

10. Teh, S.-Y., Lin, R., Hung, L.-H. & Lee, A. P. Droplet microuidics. Lab Chip 8, 198220 (2008).11. Theberge, A. B. et al. Microdroplets in Microuidics: An Evolving Platform for Discoveries in Chemistry and Biology. Angew. Chem. Int. Edit. 49, 58465868 (2010).

12. Baret, J.-C. et al. Fluorescence-activated droplet sorting (FADS): efficient microuidic cell sorting based on enzymatic activity. Lab Chip 9, 18501858 (2009).

13. Romero, P. A., Tran, T. M. & Abate, A. R. Dissecting enzyme function with microuidic-based deep mutational scanning. Proc. Natl. Acad. Sci. USA 112, 71597164 (2015).

14. Kintses, B., Hein, C., Mohamed, M. F., Fischlechner, M., Courtois, F., Lain, C., & Hollfelder, F. Picoliter cell lysate assays in microuidic droplet compartments for directed enzyme evolution. Chem. Biol. 19, 10011009 (2012).

15. Agresti, J. J. et al. Ultrahigh-throughput screening in drop-based microuidics for directed evolution. Proc. Natl. Acad. Sci. USA 107, 40044009 (2010).

SCIENTIFIC REPORTS

9

www.nature.com/scientificreports/

16. Ostafe, F., Prodanovic, R., Lloyd, Ung, Weitz, D. A. & Fischer, R. A. high-throughput cellulase screening system based on droplet microuidics. Biomicrouidics. 8, 041102 (2014).

17. Sjostrom, S. L. et al. High-throughput screening for industrial enzyme production hosts by droplet microuidics. Lab Chip 14, 806813 (2014).

18. Wang, B. L. et al. Microuidic high-throughput culturing of single cells for selection based on extracellular metabolite production or consumption. Nature Biotechnol. 32, 473478 (2014).

19. El Debs, B., Utharala, R., Balyasnikova, I. V., Griffiths, A. D. & Merten, C. A. Functional single-cell hybridoma screening using droplet-based microuidics. Proc. Natl. Acad. Sci. USA 109, 1157011575 (2012).

20. Brand, A. & Gow, N. A. R. Mechanisms of hypha orientation of fungi. Curr. Opin. Microbiol. 12, 350357 (2009).21. Khalaj, V., Brookman, J. L. & Robson, G. D. A Study of the Protein Secretory Pathway of Aspergillus niger Using a GlucoamylaseGFP Fusion Protein. Fungal Genet. Biol. 32, 5565 (2001).

22. Anna, S. L., Bontoux, N. & Stone, H. A. Formation of dispersions using ow focusing in microchannels. Appl. Phys. Lett. 82,

364366 (2003).

23. Holtze, C. et al. Biocompatible surfactants for water-in-uorocarbon emulsions. Lab Chip 8, 16321639 (2008).24. Ahn, K. et al. Dielectrophoretic manipulation of drops for high-speed microuidic sorting devices. Appl. Phys. Lett. 88, 024104 (2006).25. Ramos, A., Morgan, H., Green, N. G. & Castellanos, A. Ac electrokinetics: a review of forces in microelectrode structures. J. Phys. D Appl. Phys. 31, 23382353 (1998).

26. Sherwood, J. D. Breakup of uid droplets in electric and magnetic elds. J. Fluid Mech. 188, 133146 (1988).27. Lowe, K. C., Davey, M. R. & Power, J. Peruorochemicals: their applications and benets to cell culture. Trends in Biotechnol. 16, 272277 (1998).

28. Clausell-Tormos, J. et al. Droplet-Based Microuidic Platforms for the Encapsulation and Screening of Mammalian Cells and Multicellular Organisms. Chem. Biol. 15, 427437 (2008).

29. Chabert, M., Dorfman, K. D. & Viovy, J.-L. Droplet fusion by alternating current (AC) eld electrocoalescence in microchannels. Electrophoresis 26, 37063715 (2005).

30. Abate, A. R., Hung, T., Mary, P., Agresti, J. J. & Weitz, D. A. High-throughput injection with microuidics using picoinjectors. Proc. Natl. Acad. Sci. USA 107, 1916319166 (2010).

31. Sciambi, A. & Abate, A. R. Accurate microuidic sorting of droplets at 30 kHz. Lab Chip 15, 4751 (2015).32. Studer, A. et al. Fluorous Synthesis: A Fluorous-Phase Strategy for Improving Separation Efficiency in Organic Synthesis. Science 275, 823826 (1997).

33. Skhiri, Y. et al. Dynamics of molecular transport by surfactants in emulsions. So Matter 8, 1061810627 (2012).34. Najah, M. et al. New glycosidase substrates for droplet-based microuidic screening. Anal. Chem. 85, 98079814 (2013).35. Najah, M. et al. Droplet-Based Microuidics Platform for Ultra-High-Throughput Bioprospecting of Cellulolytic Microorganisms. Chem. Biol. 21, 17221732 (2014).

36. Boitard, L. et al. Monitoring single-cell bioenergetics via the coarsening of emulsion droplets. Proc. Natl. Acad. Sci. USA 109, 71817186 (2012).

37. Joensson, H. N., Uhlen, M. & Svahn, H. A. Droplet size based separation by deterministic lateral displacement-separating droplets by cell-induced shrinking. Lab Chip 11, 13051310 (2011).

38. Ji, J. et al. Proteolysis in microuidic droplets: an approach to interface protein separation and peptide mass spectrometry. Lab Chip 12, 26252629 (2012).

39. Kster, S. K. et al. Interfacing Droplet Microuidics with Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry: Label-Free Content Analysis of Single Droplets. Anal. Chem. 85, 12851289 (2013).

40. Smith, C. A. et al. Sensitive, High Throughput Detection of Proteins in Individual, Surfactant-Stabilized Picoliter Droplets Using Nanoelectrospray Ionization Mass Spectrometry. Anal. Chem. 85, 38123816 (2013).

41. Bailey, J. E. et al. Inverse metabolic engineering: A strategy for directed genetic engineering of useful phenotypes. Biotechnol. Bioeng. 79, 568579 (2002).

42. Beneyton, T., Coldren, F., Baret, J.-C., Griffiths, A. D. & Taly, V. CotA laccase: high-throughput manipulation and analysis of recombinant enzyme libraries expressed in E. coli using droplet-based microuidics. Analyst 139, 33143323 (2014).

43. Chen, C.-H. et al. Enhancing Protease Activity Assay in Droplet-Based Microuidics Using a Biomolecule Concentrator. J. Am. Chem. Soc. 133, 1036810371 (2011).

44. Mazutis, L. et al. Droplet-Based Microuidic Systems for High-Throughput Single DNA Molecule Isothermal Amplication and Analysis. Anal. Chem. 81, 48134821 (2009).

45. Schmitz, C. H. J., Rowat, A. C., Koster, S. & Weitz, D. A. Dropspots: a picoliter array in a microuidic device. Lab Chip 9, 4449 (2009).

We thank Jean-Christophe Baret for help with data analysis and helpful discussions. This work was supported by Bpifrance and Ets. J. Soufflet (OSIRIS project).

T.B., I.P.M.W., A.D.G. and A.D. conceived the experiments. T.B. and I.P.M.W. conducted the experiments, collected the data and analyzed the results. P.P. performed some of the microuidics experiments, constructed the fungi librairies and performed the ask fermentation experiment. M.N. designed the microuidic dropmaker and perfomed the phenotypic distribution analysis. P.L. performed the robotic fungi picking. A.C. performed the robotic microtiterplate -amylase titration. E.M. synthesized the surfactant. A.D.G. and A.D. conceived the project and supervised the experiments. T.B., I.P.M.W., A.D. and A.D.G. wrote the manuscript.

Supplementary information accompanies this paper at http://www.nature.com/srep

Competing nancial interests: The authors declare no competing nancial interests.

How to cite this article: Beneyton, T. et al. High-throughput screening of lamentous fungi using nanoliter-range droplet-based microuidics. Sci. Rep. 6, 27223; doi: 10.1038/srep27223 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

SCIENTIFIC REPORTS

10