FIGURE 2.

Droplet-based microfluidic selection strategy. (A) Droplet-based microfluidic workflow. Gray and white boxes distinguish on-chip from off-chip steps, respectively. (B–I) A typical selection is depicted and the analysis profiles corresponding to subsets of 15,000 (F,G) or 65,000 (H,I) droplets are shown. A gene library was diluted in PCR reagents to give an average, λ, of 0.2 DNA molecules per droplet. The mixture was injected into a droplet generator (B) and compartmentalized in 2.5 pL droplets by hydrodynamic focusing of the aqueous phase (dark orange) with an oil phase (gray) containing fluorosurfactant (Anna et al. 2003). The presence of 1 mg/mL Dextran-Texas Red 70 kDa (orange fluorescence) allowed droplet production to be monitored (population 1 in F). After thermocycling each 2.5 pL PCR droplet was reinjected into a fusion device (C) and paired with 16 pL droplets generated on-chip and containing an in vitro transcription system and 20 µg/mL Dextran-Texas Red 70 kDa Texas Red (orange). Pairs of droplets were electrocoalesced by passing between a pair of electrodes (350 V) and fusion efficiency was estimated on-line by measuring orange fluorescence (G). Nonfused 16 pL droplets were characterized by low orange fluorescence (population 2), while those fused with 1 (population 3) or 2 (population 4) 2.5 pL PCR droplets had a significantly increased orange fluorescence. After incubation, droplets were reinjected into a chip to allow picoinjection of the fluorogenic substrate and 7-amino-4-methyl-3-coumarinylacetic acid (blue fluorescent) (D). Droplets were spaced with oil and 4 pL of the substrate/coumarin mixture was injected into each droplet with a 150 V AC field applied. The efficiency of the process was assessed by monitoring the blue fluorescence of the coumarin (H). Populations 5, 6, and 7 correspond to populations 2, 3, and 4 in G, respectively. Finally the emulsion was reinjected into a microfluidic droplet sorter (E). Blue fluorescence was used to detect picoinjected droplets and green fluorescence to measure the products of the reaction catalyzed by the ribozyme variant present in the droplet. Droplets that contained active ribozymes (green in E, population 9 in I) were gated and deflected into the sort channel (top exit channel on inset pictures) by applying a 1200 V AC pulse. All other droplets were allowed to flow into the waste channel (bottom exit channel on inset pictures). Note that the lower blue fluorescence value on I (compared with H) is due to the use of a lower magnification objective for the sorting step. For each device, the zone where the fluorescence measurement was made is indicated by a blue arrow. Blue lines on micrographs E show the location of the laser lines used to scan the droplets.