DETERMINISTIC LATERAL DISPLACEMENT DEVICE FOR DROPLET

SEPARATION BY SIZE – TOWARDS RAPID CLONAL SELECTION BASED

ON DROPLET SHRINKING Haakan N Joensson*, Mathias Uhlén and Helene Andersson Svahn

Royal Institute of Technology (KTH), SWEDEN

ABSTRACT

We present a novel method for robust passive separation of microfluidic droplets by size using deterministic lateral

displacement(DLD). We also show that droplets containing Saccharomyces Cervisiae shrink significantly during incu-

bation while droplets containing only yeast media retain their size. We demonstrate the DLD device by sorting out

shrunken yeast-cell containing droplets from a 40-fold excess of ~33% larger yeast-cell-free droplets generated at the

same time, suggesting that DLD might be used for clonal selection. The same device also separates 11μm from 30μm

droplets at a rate of 12000droplets/second, more than twofold faster than previously demonstrated passive hydrodynamic

separation devices [1].

KEYWORDS: Droplets, Separation, High-Throughput, Cells

INTRODUCTION

Droplet microfluidics is increasingly being used in complex manipulations, which can yield polydisperse emulsions.

Subsequently regaining monodisperse emulsions after droplet fusion, heating induced coalescence in e.g. PCR or other

droplet manipulations resulting in polydisperse emulsions can be useful or necessary for further downstream processing.

In a recent publication this issue has been addressed through hydrodynamic passive separation in a trident channel at

rates <4500Hz [1]. Here we apply the DLD concept to this problem, increasing throughput more than twofold and at the

same time demonstrating the possibility of using DLD to separate highly deformable particles, i.e. microfluidically gen-

erated droplets. DLD is a high throughput size separation method which has been shown to separate solid micrometer

sized particles with a tolerance of less than 10 nm [2] utilizing a tilted pillar array,. In DLD devices, objects are sepa-

rated depending on a critical diameter (DC) which depends on the design parameters of the array, shown in Figure 1 and

described by the equation DC=2 Gd/w, where is the parabolic flow correction factor.

S.Cervisiae is a common biotechnological model system and is used in bioproduction of e.g. fuel ethanol [3] and for

production of specific biomolecules, such as the antimalarial drug precursor artemisinic acid [4], using strains specifi-

cally engineered by synthetic biology. One approach to selecting novel yeast strains with specific biological properties is

through directed evolution, recently demonstrated in microfluidic droplets using a fluorogenic substrate and serial dielec-

trophoretic sorting to select the strains producing the most efficient enzyme from a genetic library at high throughput [5].

Figure 1: Illustration showing the mechanism for DLD separation by size. In a channel with solid pillars (shown in

grey) separated by a width w, having a gap width G and a row shift distance d, objects smaller than a critical diameter

DC (represented by the green dot) can follow and remain in a single flow field and pass through the device in the direc-

tion of the flow while an object larger than the critical size (represented by the red dot) is forced into the next flow field

to the right when passing a pillar causing it to pass through the device at an angle in the direction of the tilt of the array.

EXPERIMENTAL

Microfluidic devices were manufactured in Polydimethylsiloxane by standard soft lithography. Aqueous droplets

were generated by pressure driven flow in a fluorinated carrier oil (HFE 7500, 3M) stabilized by a surfactant (EA, Rain-

Dance Technologies) in the oil phase. The DLD circuit can generate 11 and 30μm aqueous droplets or allow input of

pre-generated emulsions. Droplets were injected at the center of a large channel with a tilted pillar array with 60μm cy-

lindrical pillars separated by gaps of 60μm and a period of 10 rows (Figure 2). Oil was injected from the two side inlets

O1-2 at flow rates of 5ml/hr generating a laminar flow along the main channel. At the outlet reservoir (EC) a row of 9

978-0-9798064-3-8/µTAS 2010/$20©2010 CBMS 1355 14th International Conference onMiniaturized Systems for Chemistry and Life Sciences

3 - 7 October 2010, Groningen, The Netherlands

outlets arranged perpendicular to the flow direction connected to emulsion collection vials. The resulting device has a

theoretical critical diameter for separation of 24μm (for =2). The depth of all channels was 30μm. 50μm droplets

containing on average two and 30μm droplets containing on average 0.03 S. Cervisiae cells in YPD medium were

generated in a separate device and incubated for 8 or 36 hours respectively at 30°C. After incubation most of the 30μm

yeast/YPD emulsion was injected into the DLD device for separation while a small subset was re-injected into an

analysis circuit [6]. Here droplets were injected with additional oil and individually passed through a laser.

Fluorescence, i.e. the autofluorescence of the yeast media, was sampled using a photomultiplier tube to determine

relative droplet size.

Figure 2: Droplet size separation in the DLD device. Positions of the microscope images are approximate. A)

Streams of 11 μm and 30 μm droplets are combined in the centre input channel. B) The two populations are injected into

the pillar array. The 30 μm droplets can be observed immediately to the right of the tilted column of pillars while the 11

μm droplets are seen following the streamlines of the laminar flow through the main channel. C) The two droplet popu-

lations are separated as they move through the array. The D) 11 μm droplets and E) 30 μm droplets are completely

separated and can be collected separately at the outlet. Center Schematic) The design of the droplet DLD device with oil

inlets (O1-4) and aqueous inlets (W1-2) and outlet. Flow rates of aqueous fluid of 30μl/hr and 10μl/hr were applied at

W1 and W2, respectively; oil of 500μl/hr and 600μl/hr to O3 and O4, respectively and oil of 5ml/hr to each of O1 and

O2. Scale bar is 100μm. EI indicates an alternate emulsion inlet and EC indicates emulsion collection area.

RESULTS AND DISCUSSION

Operating the DLD chip at moderate flow rates yielded 12000 and 200 droplets per second of the 11μm and 30μm drop-

lets respectively, thus resulting in a 60-fold excess of smaller droplets. However, this bias in size distribution did not chal-

lenge the separation performance; all observed droplets were correctly separated. Upon injection into the pillar array 11μm

droplets were observed passing through the array following the direction of the flow while 30μm droplets were seen follow-

ing the tilt angle of the array (Figure 2) resulting in the complete separation of the droplets into two discrete monodisperse

emulsions immediately before entering the outlet reservoir.

Figure 3: Selective droplet shrinkage in yeast-containing droplets. Images of Saccharomyces Cervisiae encapsulated

in monodisperse 50μm droplets taken A) directly following encapsulation and B) after 8 hours of incubation. Droplets

containing yeast cells were observed to shrink by different but significant amounts while droplets not containing a cell

remained the original size. Scale bar is 50μm.

The 50μm droplets were imaged before and following incubation (Figure 3). Interestingly, yeast-containing droplets

had shrunk substantially while empty droplets remained their original size. Meanwhile the yeast cells had divided.

Figure 4 shows the separation of shrunken droplets by size in a DLD device. The 30μm droplets generated from the

yeast containing aqueous phase were monodisperse upon generation but most yeast containing droplets had shrunk to on

average 22.5μm following incubation. The ratio of shrunk droplets to those retaining their original size was measured by

laser induced fluorescence to be 2.5/100. From this starting material shrunken droplets were concentrated and collected

at a high purity (Figure 4G).

1356

Figure 4: Separation of shrunken yeast droplets. A) 30μm droplets containing YPD media and a yeast concentration

of ~0,03 cells/droplet were generated. Droplets were incubated for 36 hrs at 30°C. B) Droplet width distribution as-

sayed by laser induced fluorescence of YPD media filled droplets (ex 491nm; em 515/15nm; 15000 droplets) showing

2.5% shrunken droplets. C) Injection of the incubated emulsion into the device at EI (figure 3) at 50μl/hr. D) Droplets

not containing (30um) and E) containing yeast cells (avg 22.5μm) are separated in the device and simultaneously col-

lected in one of 7 different tubes at the outlet reservoir (EO, figure 3). F) and G) shows the collected fractions from the

two centrally placed outlets, revealing concentration of yeast containing droplets in G). Scale bar is 100μm.

CONCLUSION

We demonstrate high throughput separation of microfluidically generated droplets in a DLD device at a rate of 12000

droplets per second. We also observe the selective shrinking of yeast containing droplets and separate these droplets by

size based on this biological determining factor, separating a droplet population with a mean diameter of 22.5μm from a

population of 33% larger diameter present at a 40-fold excess. We are now characterizing the device more precisely us-

ing higher separation rates as well as investigating the efficient selection of yeast strains based on biological characteris-

tics by separating shrunken droplets.

ACKNOWLEDGEMENTS

This work was funded by the Swedish Research Council, the Knut and Alice Wallenberg Foundation and VINNOVA

through the ProNova VINN Excellence Center. The surfactant used was graciously provided by RainDance Technolo-

gies, Boston, MA, USA.

REFERENCES

[1] L. Matuzis and A. Griffiths Preparation of monodisperse emulsions by hydrodynamic size fractionation, Appl.

Phys. Lett., 95, 204103 (2009)

[2] L.R. Huang, E.C. Cox, R.H. Austin and J.C. Sturm, Continuous lateral displacement Through Deterministic Lat-

eral Displacement, Science, 304, 987 (2004)

[3] S. Banerjee, S. Mudliar, R. Sen, B. Giri, D. Satpute, T. Chakrabarti and R.A. Pandey, Commercializing lignocellu-

losic bioethanol: technology bottlenecks and possible remedies, Biofuels, Bioprod. Bioref. 4, 77 (2010)

[4] D-K. Ro, E. M. Paradise, M. Ouellet, K. J. Fisher, K. L. Newman, J. M. Ndungu, K. A. Ho, R. A. Eachus, T. S.

Ham, J. Kirby, M. C. Y. Chang, S. T. Withers, Y. Shiba, R. Sarpong, and J. D. Keasling., Production of the antima-

larial drug precursor artemisinic acid in engineered yeast. Nature 440 940-943 (2006)

[5] J.J. Agresti, E. Antipov, A.R. Abate, K. Ahn, A.C. Rowat, J-C Baret, M. Marquez, A.M. Klibanov, A.D. Griffiths,

and D.A. Weitz, Ultrahigh-throughput screening in drop-based microfluidics for directed evolution, PNAS, 107, 9

(2010)

[6] H.N. Joensson, M.L. Samuels, E.R. Brouzes, M Medkova, M. Uhlén, D.R. Link and H. Andersson-Svahn, Detec-

tion and Analysis of Low-Abundance Cell-Surface Biomarkers Using Enzymatic Amplification in Microfluidic

Droplets, Angew Chem Int Ed, 48, 14, (2009)

CONTACT

*H.N. Joensson, tel: +46-8-5537-8345; hakan.jonsson@biotech.kth.se

1357