This journal is©The Royal Society of Chemistry 2017 Integr. Biol., 2017, 9, 687--694 | 687

Cite this: Integr. Biol., 2017,9, 687

Demonstration of transgressive overyielding ofalgal mixed cultures in microdroplets†

David N. Carruthers, ‡ab Chang Kyu Byun,‡ac Bradley J. Cardinaleb andXiaoxia Nina Lin*a

Algae are ubiquitous in natural ecosystems and have been studied extensively for biofuel production due

to their unique metabolic capabilities. Most studies to date have approached biofuel optimization

through synthetic biology and process engineering with few industrial scale projects considering algal

community interactions. Such interactions can potentially lead to increased productivity and reduced

community invasability, both important characteristics for scalable algal biofuel production. It is

estimated that over a million species of algae exist such that elucidating the interactions that might be

beneficial for biofuel production remains extremely resource and time intensive. Here we describe a

strategy for rapid, high-throughput screening of algal community combinations using a microfluidic

platform to generate millions of parallel, nanoliter-scale algal mixed cultures for estimation of biomass

accumulation. Model communities were first studied in a bench scale flask experiment and then

examined using microfluidic droplets. These experiments showed consistent results for both positively

interacting algal bicultures that increase biomass when together, and negatively interacting bicultures

that decrease biomass. Specifically, these included enhanced performance of two bicultures,

Ankistrodesmus falcatus and Chlorella sorokiniana, Chlorella sorokiniana and Selenastrum minutum, and

reduced performance of a biculture consisting of Selenastrum capricornutum and Scenedesmus

ecornis. While the ultimate techno-economic feasibility of algal bioproducts hinges on an amalgamation

of scientific fields, rapid screening of algal communities will prove imperative for efficiently discovering

community interactions.

Insight, innovation, integrationSynergy within microbial communities can lead to increased productivity and reduced community invasability, which are beneficial characteristics to scalablebioproduction platforms. Rapid, high-throughput screening of microbial combinations using microfluidic devices as demonstrated in this work could proveimperative for efficient and economical discovery of such synergistic communities.

Introduction

Microalgae show promise in their viability for the production ofcleaner, renewable sources of energy with respect to their fossilfuel analogues. They also provide an attractive opportunity for

industrial development due to the inherent compatibility ofalgal lipid derived fuels with the current petroleum-based infra-structure. However, the methods with which algae may beengineered competitively for market operation remain especiallychallenging. A dominant approach in algal biofuel synthesisinvolves the development of genetically modified organisms(GMOs) to optimize laboratory yields.1 Researchers have demon-strated the ability to manipulate seemingly innocuous targetmicrobes to create a product with industrial viability and thepotential to meet market demands through metabolic engineer-ing and increased lipid production. Novel processing techniquessuch as the hydrothermal liquefaction of algal biomass havecomplemented these advances with the production of up to 64%biocrude per unit biomass.2

a Chemical Engineering Department, University of Michigan, Ann Arbor, Michigan,

USA. E-mail: ninalin@umich.edub School of Natural Resources and the Environment, University of Michigan,

Ann Arbor, Michigan, USAc Department of Applied Chemistry, DaeJeon University, Daejeon,

The Republic of Korea

† Electronic supplementary information (ESI) available: Supplementary informa-tion 1 and 2. See DOI: 10.1039/c6ib00241b‡ Equal contribution.

Received 16th December 2016,Accepted 12th June 2017

DOI: 10.1039/c6ib00241b

rsc.li/integrative-biology

Integrative Biology

PAPER

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article OnlineView Journal | View Issue

688 | Integr. Biol., 2017, 9, 687--694 This journal is©The Royal Society of Chemistry 2017

Nonetheless the dominant approach has been hampered by avariety of problems, including horizontal gene transfer in whichspecies naturally evolve and change their properties, functionaltrade-offs in metabolism that make it difficult to simultaneouslyoptimize multiple fuel properties, and the inherent instability ofmonoculture cultivation with respect to disturbances andinvaders.3,4 These effects have been identified as major barriersfor maximizing the long-term productivity and stability of openpond biofuel systems.5 A multifaceted approach is required toaddress the enormity of this challenge. Engineering specificstrains for production capacity, while useful, cannot be consid-ered as the standalone approach to algal biofuel research. Anapproach to reduce the disparity between laboratory and outdoorcultivation must integrate biological and process engineeringdevelopments within an ecological context.

Ecological engineering is an emergent approach in algalbiofuel optimization that addresses limitations in yield andstability through the selection of algal consortia with a pre-disposition for increasing productivity. Such productivity isintimately linked to interspecific interactions involving compe-tition, niche partitioning, or facilitation. Competition involvesorganisms contesting the same limited resources, leading to achange in fitness that negatively impacts one or both organisms.Niche partitioning occurs when species are limited by differentresources, or divide the same resource in space or time.6 Completeniche partitioning indicates the absence of interspecific inter-actions altogether such that individual species perform equallywell in monoculture as in polyculture. Facilitation is a positiveinteraction in which at least one of the constituent organismswithin the community benefits, quantified by an increase inoverall production. If an interspecific interaction is sufficientlystrong, then a net change in biomass may be substantiallymeasured through transgressive overyielding (TO).7 In this con-text, TO occurs when the biomass production of a community ofalgae species exceeds that of the most productive constitutivespecies. The degree to which TO arises is dependent on thenature of the community, specifically the relative carrying capa-cities and growth rates of subspecies.8 In a meta-analysis of plantspecies, Cardinale et al. showed that polycultures produce anaverage of 1.7 times more biomass than their constituent speciesgrown in monoculture, with increased productivity in 79% of 96polyculture experiments.7 However, only 12% of these polycul-tures demonstrated TO with respect to the most productiveconstitutive monoculture and the ratio of polyculture to consti-tutive highest producing monoculture was 0.87 on average.7 Inalgal polycultures, Shurin et al. conducted an experiment with119 species pairs, of which 25% produced more biomass thanthe average monoculture.9 In these experiments, the ratio ofpolyculture biomass to constitutive highest producing mono-culture was 0.82 on average.9 While TO relationships certainlyarise in algal polycultures, combinations that generate morebiomass with respect to the dominant species are elusive. Theprevalence of TO relationships is further convoluted by contra-dictory research and varied experimental conditions consideringthe effects of species richness on biomass and biocrude yield.10,11

While tending to overlook the mechanisms through which these

interactions occur (metabolite exchange, niche partitioning,microbial symbionts, etc.), these studies are nonetheless pro-mising in their relation to scaled algal production. Algal pondstypically facilitate a single species in its fundamental niche,relatively independent of interspecific interactions. The singu-larity of this approach thus ignores potentially lucrative com-binatorial overyielding relationships. Yet prospecting for theseoveryielding relationships can be tedious, expensive, and extre-mely time consuming particularly given that over 44 000 species ofalgae have currently been identified with a maximum of 1 millionmore estimated to exist.12,13 Studies have been limited in practiceto several hundred combinations. Such interactions are alsodifficult to model and predict due to the limiting density depen-dence or resource provisions of the individual species.8,9 Thediversity of algal species necessitates the development of a high-throughput platform by which algal polycultures can be quicklyanalyzed, with relationships elucidated and considered on thebasis of productivity (e.g. measured by biomass accumulation).Selected species combinations could then be considered ascandidates for thermochemical processing, by which the entirealgal biomass is converted into a bio-crude product.14

Algal experiments in the laboratory are conventionally con-ducted at the bench scale, consuming large quantities of resources.At this scale, it is impossible to screen tens of thousands ofspecies combinations for overyielding. Screening of algal com-munities might, therefore, benefit from microfluidic dropletgenerators (MFDG), which have been ameliorated and incorpo-rated within a broad range of research topics. Aqueous dropletsseparated by a surrounding continuous medium form parallelindependent mini-compartments, providing an effective meansfor high-throughput screening.15 Additionally, the miniaturiza-tion of fluidic systems facilitates conservation of reagents, experi-mental expenses, and increases general efficiency.16 Tunabilityand controllability of each droplet allows for parallel processingsuch that large datasets are quickly and easily generated withextremely high resolution.16,17 Studies have succeeded in encap-sulating and culturing various species of algal monocultures forthe study of growth kinetics with regards to pH, nitrogenconcentration, and resource consumption.18 Park et al. demon-strated mutualistic interactions in synthetic E. coli communitiesin microdroplets on a platform readily applicable to algae.19 Theaqueous phase may also be readily substituted with many targetmedia including wastewater sources to study algal polyculturegrowth for bioremediation.20 Adaptation of this platform wouldenable rapid creation of highly parallel algal communities forhigh-throughput spectroscopic analysis of the fluorescent mole-cule chlorophyll-a, which provides a metric by which growth, andhence production, may be compared between monocultures andpolycultures for the elucidation of TO.21

In this work, we have demonstrated the feasibility of usingmicrofluidic droplet co-cultivation and microscopic fluorescenceanalysis for detecting TO of algal polycultures. The overallconcept is depicted in Fig. 1A and B and a real droplet generationdevice is shown Fig. 1C. First, a set of flask experiments wasperformed to provide benchmark data. Then, a microdropletgeneration device was fabricated using the design illustrated

Paper Integrative Biology

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article Online

This journal is©The Royal Society of Chemistry 2017 Integr. Biol., 2017, 9, 687--694 | 689

in Fig. 1D. The droplet platform was characterized for algalcultivation in terms of droplet size, uniformity, stability, and celldistribution. Finally, a translation of flask experimental results todroplets validated the proposed platform for future polyculturescreening.

Results and discussionAlgal cultivation in Erlenmeyer flasks

Transgressive overyielding (TOc) can be quantified using thenatural log of the ratio of biculture biomass with respect to themost productive monoculture within the combination as follows:

TOc ¼ lnBbiculture

Bmonoculture

� �

Higher biomass in biculture will lead to a positive value fortransgressive overyielding whereas higher biomass in mono-culture yields a negative value. Algal cultures were initiallystudied in a flask experiment to identify overyielding combina-tions for translation into the droplet co-cultivation framework(Supplementary Information 1). Of the combinations explored,two bicultures demonstrated transgressive overyielding withstatistical significance (two sample one-tailed Student’s t-test).These bicultures were A. falcatus and C. sorokiniana (p = 0.0014)and C. sorokiniana and S. minutum (p = 0.0004), which demon-strated TO values of 0.59 � 0.1 and 0.59 � 0.09, respectively.Given the logarithmic nature of TO, this means the biculturesachieved approximately 1.8 times the biomass of the mostproductive monoculture. A third biculture of particular interestwas S. ecornis and S. capricornutum, which demonstrated astatistically significant underperformance (p = 0.014) with aTO value of �0.42 � 0.41 or 0.66 times the biomass of the mostproductive monoculture. These bicultures were then used asmodel systems for the validation of the microfluidic dropletco-cultivation framework.

Microdroplet generation and incubation

Fig. 1A illustrates the set-up for droplet generation and storage.A PFTE tube (1.5 mm inner diameter) was used to connect the

droplet generation device to a sterile 1.7 mL Eppendorf tube.After generation, droplets were stored within the Eppendorftube with the addition of generic mineral oil. As a less densefluid, the mineral oil provided an insulating layer above themicrodroplets, avoiding evaporative losses. Microdroplets restedabove excess oil used in droplet generation. This method pro-vided a sterile environment for bulk droplet storage from whichdroplets were extracted and visualized on disposable hemocyto-meters (iNCYTO C-Chip DNC-N01) for point reads over the algaegrowth cycle.

Droplet encapsulation of microbes has been previouslydemonstrated to follow a Poisson distribution.19 An initialstudy in this work considered monocultures of A. falcatus (A),C. sorokiniana (C) and a biculture consisting of these two species(AC). Biculture feedstock was made by mixing both speciestogether in a 1 : 1 ratio at a specified cell density, avoiding thenecessity of multiplying distributions from independent aqueousstreams combining in the device.22 A biculture feedstock inher-ently produces heterogeneous droplets such that some dropletscontain monocultures of A, some contain monocultures of C, andothers contain bicultures of AC. The selected l value must thusbe large enough to facilitate the generation of all three possibi-lities and not large enough to generate only bicultures to facilitateinterdroplet comparison from a single feedstock. Fig. 2A showsthe distribution of cells per droplet with these monocultures andbiculture. The droplets were highly uniform, as demonstrated inFig. 2B. Moreover, they were very stable, maintained over thecourse of the algal growth cycle for 24 days with a recordedaverage coefficient of variation of 0.030 � 0.004 (Fig. 2C, Supple-mentary Information 2). Generation and maintenance of uniformdroplets is critical for interdroplet comparison due to effects oncarrying capacity.18

The effects of phase flow rates on droplet volumes were alsoexamined. These quantities were previously reported to berelated proportionally in flow focusing devices.23 Droplet wereassumed cylindrical due to significant differences in channeldepth and droplet width. The characteristic orifice and channeldimensions facilitated the generation of droplets between 70 and150 mm diameter. The flow rates for 70 mm diameter droplets

Fig. 1 Microfluidic droplet co-cultivation framework and device. (A) A tube for aqueous algal culture and a tube for continuous phase enter thepolydimethylsiloxane (PDMS) device for droplet generation. Droplets are transported to a 1.7 mL Eppendorf tube with a syringe tip for sampling. (B) Themicrodroplets (2) are cultivated and stored above a layer of excess HFE oil (3) and below a layer of mineral oil (1) (not to scale). (B) Illustration oftransgressive overyielding of a biculture in droplets. (C) Picture of a flow focusing droplet generator with one-cent coin for scale. (D) CAD schematic ofmicrodroplet generator from this work.

Integrative Biology Paper

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article Online

690 | Integr. Biol., 2017, 9, 687--694 This journal is©The Royal Society of Chemistry 2017

were Foil/Faqueous = 1.4/1.6 mL min�1. Microfluidic devices with250 mm (for oil) and 200 mm (for aqueous cell suspension) inletwidths connected to 150 mm orifice widths, were used for largerdroplet of 250 and 500 mm diameters. The flow rate for 500 mmdiameter droplets was Foil/Faqueous = 3.3/5.4 mL min�1. It wasobserved that as the droplet diameter (thus volume) expanded,the maximal number of algae cells each droplet could sustainalso increased (Fig. 2D).

Algal cultivation in microdroplets

Droplets were stored under identical growth conditions as flasks.Growth curves were constructed using fluorescent point reads.A. falcatus and C. sorokiniana (AC), as well as C. sorokiniana andS. minutum (CE), were used as model systems that exhibited highlevel of transgressive overyielding in flasks. The bicultureS. ecornis and S. capricornutum (JL) was selected to representan apparent negative relationship. Each biculture study con-sisted of three Eppendorf tubes, one from a 1 : 1 biculturefeedstock and two with its respective monocultures as controls.Theoretically, each experiment requires only a single biculture

feedstock to generate all possibilities, however C. sorokinianaand S. minutum are morphologically indistinguishable underthe microscope at 10� magnification. Further magnificationbecomes analytically challenging. Therefore, bicultures werelumped in bulk and compared to individual monocultures.While flask experimental fluorescence was determined usingan M5 spectrophotometer, droplet fluorescence was calculatedand normalized as a product of pixel intensity and fluorescentarea using ImageJ.

As summarized in Fig. 3, AC biculture droplets outperformedmonoculture droplets containing A or C only, which was consis-tent with results from flask experiments (Fig. 3A, TO = 0.59� 0.1).All droplets from the biculture feedstock were regarded as con-taining bicultures and transgressive overyielding was estimatedto be TO = 0.55 � 0.09 (Fig. 3B). Alternatively, when A and Cwere distinguished based on morphological differences, it wasestimated that TO = 0.26� 0.05 (Supplementary Information 1).This lower TO estimate based on morphological distinctionwas repeatedly observed for the A. falcatus and C. sorokinianabiculture. Reduced cell growth was noted when comparing

Fig. 2 Validation of microalgae encapsulation and growth in droplets. (A) Poisson distribution of cells per droplet for A. falcatus, C. sorokiniana, and thebiculture feedstocks after droplet generation (n = 50). (B) Image of droplets of C. sorokiniana at day 0 showing high degree of monodispersity.(C) Diameter of C. sorokiniana droplets during algal growth demonstrating sustained storage in droplets (n = 10). (D) Maximal cell numbers of C. vulgarisin droplets of different volumes (n = 3).

Paper Integrative Biology

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article Online

This journal is©The Royal Society of Chemistry 2017 Integr. Biol., 2017, 9, 687--694 | 691

A. falcatus monoculture droplets in biculture and strict mono-culture conditions. One possible explanation is that bulkbiculture cultivation negatively affected A. falcatus cell growthbecause of the presence of the more abundant C. sorokiniana,reducing light availability. This observation should be examinedmore carefully in a future investigation to determine any plausibleinterdroplet interactions.

A subsequent study involved C. sorokiniana (C) and S. minutum.(E), which are morphologically indistinguishable in this work.Fig. 4B indicates that the droplet trends mimic those of the flasks(TO = 0.58 � 0.09), Fig. 4A, with the biculture reaching a higherfluorescence even with the underestimation (TO = 0.29 � 0.14)and thus indicating higher biomass accumulation in the biculturewith respect to the most productive monoculture.

Lastly, S. ecornis and S. capricornutum were examined tostudy whether this analysis could also distinguish negativerelationships. In Fig. 4D, S. capricornutum outperforms the

combined biculture in droplets over their respective growth cycles(TO = �0.42 � 0.30). Despite a poor logistic fit, this trend was ingood agreement with the flask experiment (TO = �0.22 � 0.11)depicted in Fig. 4C, indicating lower biomass accumulation withinthe biculture. It should be noted that as in the CE experiment, thetwo species J and L were not distinguished morphologically here.

The three droplet experiments described above validate themicrofluidic platform in its ability to identify transgressive over-yielding of algal co-cultures. In the meantime, this work suggestsan analytical bottleneck that hinders application to more com-plex polycultures if the identification of TO combinations indroplets would rely on morphological differences betweenspecies. Another issue observed is algal quiescence where cellscontinue to fluoresce but fail to reproduce. While the micro-droplet co-cultivation method described here provides a foun-dation for high-throughput screening, algal combinationsdemonstrating TO in droplets would need to be further studied.In addition to more accurate TO quantification, these follow-upexperiments would also reveal underlying mechanisms such asniche partitioning and metabolic cross-feeding.

ExperimentalAlgal species selection

Species selected from the Cardinale Lab’s algal library andpreviously investigated in a mesocosm experiment24 wereutilized in this study. Experiments were conducted using BOLD3 N liquid medium, which is common for freshwater xenicmicroalgal cultures. The BOLD 3N medium contains 8.82 mMNaNO3, 0.17 mM CaCl2�H2O, 0.30 mM MgSO4�7H2O, 0.43 mMK2HPO4, 1.29 mM KH2PO4, and 0.49 mM NaCl. Additionally, aP-IV metal solution is used to provide 2.01 mM Na2EDTA�2H2O,0.36 mM FeCl3�6H2O, 0.21 mM MnCl2�4H2O, 0.037 mM ZnCl2,0.0084 mM CoCl2�6H2O, and 0.017 mM Na2MoO4�2H2O. Of these,sodium nitrate and potassium phosphate are the sole sourcesof nitrogen and phosphorus with atmospheric CO2 as the solecarbon source.

Seven species, Ankistrodesmus falcatus, Pediastrum duplex,Clorella sorokiniana, Scenedesmus acuminatus, Scenedesmus ecornis,Selenastrum capricornutum, and Selenastrum minutum were culti-vated as monocultures and selected bicultures in quadruplicateflask cultures at an initial concentration of 1 � 104 cells per mLin 50 mL BOLD 3N medium. These were shaken at 120 RPMunder constant luminous flux of 1375 lumens at 25 1C in arefrigerated incubator (Isotemp 2000, ThermoFisher ScientificInc. Fife, Washington, USA). In terms of photosynthetically activeradiation, light intensity was measured to be approximately100 mmol photons m�2 s�1 (Spectrum Technologies, Inc). Dailyfluorescence readings were taken using 200 mL sample aliquotsover a period of approximately 45 days on a Spectramax M5spectrophotometer (Molecular Devices). Every five days 1 mL offresh medium was added to account for culture volume reduc-tion with evaporative losses assumed negligible. Flasks wereremoved after 2 or 3 days of declining fluorescence or signifi-cant changes in colour for measurement of biomass in the

Fig. 3 A. falcatus and C. sorokiniana experiments in flasks and micro-droplets. (A) Comparison of growth curves of A. falcatus and C sorokinianamonocultures and biculutre in flasks over their respective growth cycles(n = 4). (B) Comparison of monocultures and biculture in droplets (n = 20).Flask error bars represent standard deviation while droplet error bars repre-sents standard error. Curves represent a best fit logistic regression for eachexperiment.

Integrative Biology Paper

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article Online

692 | Integr. Biol., 2017, 9, 687--694 This journal is©The Royal Society of Chemistry 2017

stationary phase. Fluorescence data were fit to the Verhulstlogistic growth model through the MATLAB curve fitting pro-gram. It is worth noting that in general the biculture experi-ments exhibited biphasic growth curves, causing worse fits ofthe model. This likely arose in large part due to the complexinteractions and dynamics in these bicultures. The carryingcapacity and growth rate for each culture were determined withfluorescence based carrying capacities used in the calculation oftransgressive overyielding. A similar strategy was employed fordroplet analysis. The overall shapes of growth curves in dropletsdo not exactly resemble those in flasks due to the smallernumber of time points and larger variations from the inherentstochasticity of small cell numbers in droplets. Nevertheless,logistic fits tended to fall within acceptable confidence intervalswith some variability (Supplementary Information 1). A compar-ison of fluorescence and dry weight biomass was conducted bycentrifugation, removal of supernatant, and oven drying at 55 1Cfor 72 hours per conical flask. This allowed for physical measure-ment of endpoint biomass of each culture at their estimatedcarrying capacity. Selected species combinations AC, CE, and JL

demonstrated similar relationships between fluorescence andbiomass overyielding, asserting the validity of chlorophyll-afluorescence as a proxy for biomass. Explicitly, these biomassTO values were 0.47� 0.15, 0.38� 0.38, and�1.51� 1.67 for AC,CE, and JL, respectively. While chlorophyll-a has some limita-tions as a pigment based method for approximating biomassaccumulation, it remains a common and realistic method ofanalysing cellular growth on the microscale. Method validationhas considered intercellular variations in chlorophyll-a concen-tration within each species and over their growth cycles. Suchintraspecial and temporal chlorophyll-a variations are well-knowncaveats when using this fluorescent proxy.21 It was found thatspecies specific chlorophyll-a fluorescence intensity did not varysignificantly intercellularly over the growth cycle, with totalfluorescence values increasing primarily due to changes in fluor-escent area (Supplementary Information 1). Furthermore, growthwas terminated at the beginning of stationary phase to avoid theinherent nonlinearity between biomass and chlorophyll-a concen-trations during stationary phase. Variations in chlorophyll-a con-centration between species were accounted for via aforementioned

Fig. 4 Comparison of bicultures in flasks and microdroplets. Growth curves of C. sorokiniana and S. minutum monocultures and their biculture in flasks(A, n = 4) and in microdroplets (B, n = 20). Growth curves of S. ecornis and S. capricornutum monocultures and their biculture in flasks (C, n = 4) and inmicrodroplets (D, n = 20). Flask error bars represent standard deviation while droplet error bars represets standard error. Curves represent a best fitlogistic regression for each experiment.

Paper Integrative Biology

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article Online

This journal is©The Royal Society of Chemistry 2017 Integr. Biol., 2017, 9, 687--694 | 693

normalization methods. Microdroplets were generated at anexpected lambda of 1.5 to account for flocculation and thecylindrical approximation for an initial concentration of3.82 � 106 cells per mL per culture. Although S. ecornis formssmall coenobia, this did not significantly affect image analysisor droplet generation. Analysis of larger coenobia, as commonwith Pediastrum sp., is possible through the generation of muchlarger microdroplets that are able to accommodate these largerparticles.

Microfabrication of microdroplet generator

The flow focusing droplet generating device was microfabricatedusing standard photolithography with a poly(dimethylsiloxane)(PDMS) substrate. Firstly, a photomask was designed usingL-Edit software (ver 12.1) and translated to a Heiderlberg mPG501 Mask Maker with a chrome in glass substrate. Channelswere designed with widths of 100 mm and an orifice width of50 mm. SU-8 2025 (MicroChem) negative photoresist was spin-coated in a uniform 50 mm layer on a 4-inch silicon wafer. Thedroplet generation design was transferred to the coated wafer byphotolithographic UV exposure with a mask aligner. The exposedwafer was then developed with SU-8 developer (MicroChem). Thewafer was then baked and developed using SU-8 developer andsubsequently washed with acetone and isobutyl alcohol. Afterdeveloping, this photomask was treated with trichloro(1,1,2,2-perfluorocytl)silane in a vacuum chamber to form a monolayeron the silicon surface, reducing surface energy and increasinghydrophobicity. A 10 : 1 (w/w) mixture of pre-polymer Sylgard 184and cross-linker (Dow corning) was made and poured onto boththe patterned wafer and a blank wafer in weigh dishes. Mixtureswere degasified in a vacuum chamber for 1 hour until all bubbleswere absent and then cured at 80 1C overnight. After curing, thePDMS layers were removed from both wafers and dishes, cut intomanageable slabs with a razor blade, and cleaned with 3MScotch tape. Inlet and outlet holes were punched to facilitateinsertion of 1.5 mm PET tubing and syringes for droplet gene-ration. The layers were treated with oxygen plasma for tenseconds per layer and bonded. Devices were baked for 6 hoursto further increase hydrophobicity.

Microdroplet generation and incubation

The droplet generation process employed dual pneumatic pumpsthat used LabView programming to control voltage and henceforced air flow rates of the respective fluid phases. The systemwas set up at a microscope station equipped with a Nikon EclipseTi-S and Nikon Intensilight C-HGHI, allowing for real timeanalysis not only of droplet generation but also of fluorescenceof chlorophyll-a within generated droplets. These pneumaticpumps were connected to a dual 2.5 mL syringe system for theaqueous and continuous phases. The aqueous phase for theseexperiments involved BOLD 3N medium as was used in the flaskexperiment, with the continuous phase consisting of 3M NovecHFE 7500 Engineered Fluid with 2% (w/w) fluorosurfactant(008-fluorosurfactant, RAN Biotechnologies, Inc.) to maintaindroplet integrity over the growth cycle. Syringes were connectedwith PTFE microtubing to Norson precision tips designed for

dispensing engineered fluid. The tips were then inserted intothe microfabricated PDMS device on a glass slide. The glass-PDMS contact was sealed with epoxy for stability.

Image collection and data analysis

A SpectraMax M5 microplate reader (Molecular Devices) with96-well microplate (Greiner) was used to monitor fluorescenceof cell suspensions in Erlenmeyer flasks. 435 nm excitation and685 nm emission wavelengths were used to detect chlorophyll-afluorescence, measured in relative fluorescence units (RFU).Microdroplet generation and cell growth over time were moni-tored with camera module (EXi blue, Q imaging) of Nikon eclipseTi microscope. Both bright and fluorescent field images weretaken with Endow GFP Longpass Emission filter. Cell counts indroplets were performed using Adobe Photoshop CS5.1. Dropletfluorescence was calculated as a product of pixel intensity andfluorescent area using ImageJ with consistent thresholds,labelled as arbitrary units (a.u.). Monoculture and biculturedroplets were visually characterized through morphologicaldifferences. For each read, a 10 mL sample of droplets was takenfrom each Eppendorf tube and added to a disposable hemo-cytometer C-Chip. Twenty random droplets of each monocultureor biculture were analyzed and averaged. The C-Chips and sampleswere discarded after imaging.

Conclusions

The work described above successfully demonstrated the transla-tion of flask-size overyielding experiments to microfluidic dropletsdisplaying both positive relationships between A. falcatus andC. sorokiniana as well as C. sorokiniana and S. minutum and anegative relationship between S. ecornis and S. capricornutum.Validation of this strategy provides a foundation for applyingmicrodroplet co-cultivation to high-throughput screening ofmassive numbers of algal combinations from a complex algallibrary. Extrapolation of this framework to higher order poly-cultures will allow for swift elucidation of productive commu-nities thereby maximizing overall biomass accumulation and,ultimately, facilitating more efficient bioproduction strategies.Given the low resource inputs, it could be similarly dispatchedto quantify algal growth on varied media such as municipal andindustrial wastewater sources for bioremediation strategies.

Two major avenues for future optimization pertain to dropletstorage and droplet analysis. Recent research has explored mono-layer droplet tracking systems by which individual algal dropletsare studied for photoautotrophic growth within a PDMS micro-droplet storage device.18,25 Creating a highly controlled, humidi-fied environment in which droplets may be arranged and storedin such a device would allow for not only increased experimentalprecision, but for rapid and predictable screening of dropletfluorescence in MATLAB. Arranging droplets in this mannerfacilitates high-throughput computational analysis. A secondopportunity involves high-throughput droplet sorting. This canbe achieved by using fluorescence-activated microfluidic dropletsorters based on total fluorescence intensity26 or the number of

Integrative Biology Paper

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article Online

694 | Integr. Biol., 2017, 9, 687--694 This journal is©The Royal Society of Chemistry 2017

fluorescent particles in a droplet.27 Alternatively, double emul-sion droplets can be employed, where the oil phase is emulsi-fied in an outer aqueous phase to generate droplets compatiblewith standard FACS instruments for drop-in sorting. In eithercase, high-throughput droplet sorting would allow automatedisolation of particularly productive communities without thelaborious bottleneck of image analysis, thereby accelerating theelucidation of interspecies interactions.

Acknowledgements

This work was conducted with funding from the University ofMichigan MCubed program and the National Science Foundation(EFRI 1332342). DNC was partially supported by the University ofMichigan’s Rackham Graduate School Summer Research Award.

References

1 D. R. Georgianna and S. P. Mayfield, Exploiting diversityand synthetic biology for the production of algal biofuels,Nature, 2012, 488(7411), 329–335.

2 P. J. Valdez, et al., Hydrothermal liquefaction of bacteriaand yeast monocultures, Energy Fuels, 2013, 28(1), 67–75.

3 P. J. Keeling and J. D. Palmer, Horizontal gene transfer ineukaryotic evolution, Nat. Rev. Genet., 2008, 9(8), 605–618.

4 K. S. McCann, The diversity–stability debate, Nature, 2000,405(6783), 228–233.

5 M. Hannon, et al., Biofuels from algae: challenges andpotential, Biofuels, 2010, 1(5), 763–784.

6 G. Hardin, The competitive exclusion principle, Science,1960, 131(3409), 1292–1297.

7 B. J. Cardinale, et al., Impacts of plant diversity on biomassproduction increase through time because of species com-plementarity, Proc. Natl. Acad. Sci. U. S. A., 2007, 104(46),18123–18128.

8 G. G. Mittelbach, Community ecology, Sunderland, Mass.,Sinauer Associates.xv, 2012, p. 400.

9 J. B. Shurin, S. Mandal and R. L. Abbott, Trait diversityenhances yield in algal biofuel assemblages, J. Appl. Ecol.,2014, 51(3), 603–611.

10 A. Narwani, et al., Power of Plankton: Effects of AlgalBiodiversity on Biocrude Production and Stability, Environ.Sci. Technol., 2016, 50(23), 13142–13150.

11 M. Stockenreiter, et al., The effect of species diversity onlipid production by micro-algal communities, J. Appl. Phycol.,2012, 24(1), 45–54.

12 T. M. Mata, A. A. Martins and N. S. Caetano, Microalgae forbiodiesel production and other applications: a review,Renewable Sustainable Energy Rev., 2010, 14(1), 217–232.

13 M. D. Guiry, How many species of algae are there?, J. Phycol.,2012, 48(5), 1057–1063.

14 D. L. Barreiro, et al., Hydrothermal liquefaction (HTL) ofmicroalgae for biofuel production: state of the art reviewand future prospects, Biomass Bioenergy, 2013, 53, 113–127.

15 D. Conchouso, et al., Simulation of a 3D Flow-FocusingCapillary-Based Droplet Generator, in COMSOL Conference,Rotterdam, Netherlands, 2013.

16 A. Huebner, et al., Microdroplets: a sea of applications?, LabChip, 2008, 8(8), 1244–1254.

17 S.-Y. Teh, et al., Droplet microfluidics, Lab Chip, 2008, 8(2),198–220.

18 J. Pan, et al., Quantitative tracking of the growth of indivi-dual algal cells in microdroplet compartments, Integr. Biol.,2011, 3(10), 1043–1051.

19 J. Park, et al., Microdroplet-enabled highly parallel co-cultivationof microbial communities, PLoS One, 2011, 6(2), e17019.

20 P. Bohutskyi, et al., Effects of inoculum size, light intensity,and dose of anaerobic digestion centrate on growth andproductivity of Chlorella and Scenedesmus microalgae andtheir poly-culture in primary and secondary wastewater,Algal Res., 2016, 19, 278–290.

21 P. Falkowski and D. A. Kiefer, Chlorophyll a fluorescencein phytoplankton: relationship to photosynthesis and bio-mass, J. Plankton Res., 1985, 7(5), 715–731.

22 T. P. Lagus and J. F. Edd, High-throughput co-encapsulationof self-ordered cell trains: cell pair interactions in micro-droplets, RSC Adv., 2013, 3(43), 20512–20522.

23 Z. Nie, et al., Emulsification in a microfluidic flow-focusingdevice: effect of the viscosities of the liquids, Microfluid.Nanofluid., 2008, 5(5), 585–594.

24 A. Narwani, A. Lashaway, D. Hietala, P. Savage and B. J.Cardinale, Effects of algal biodiversity on biocrude pro-duction and stability, Environ. Sci. Technol., 2017, 50(23),13142–13150.

25 Y. J. Sung, et al., Microdroplet photobioreactor for thephotoautotrophic culture of microalgal cells, Analyst, 2016,141(3), 989–998.

26 J.-C. Baret, et al., Fluorescence-activated droplet sorting(FADS): efficient microfluidic cell sorting based on enzy-matic activity, Lab Chip, 2009, 9(13), 1850–1858.

27 Z. Cao, et al., Droplet sorting based on the number ofencapsulated particles using a solenoid valve, Lab Chip,2013, 13(1), 171–178.

Paper Integrative Biology

Publ

ishe

d on

15

June

201

7. D

ownl

oade

d by

Uni

vers

ity o

f M

ichi

gan

Lib

rary

on

26/0

9/20

17 1

5:00

:50.

View Article Online