Droplet microfluidics for time-resolved serial crystallography

Abstract

Serial crystallography requires large numbers of microcrystals and robust strategies to rapidly apply substrates to initiate reactions in time-resolved studies. Here, we report the use of droplet miniaturization for the controlled production of uniform crystals, providing an avenue for controlled substrate addition and synchronous reaction initiation. The approach was evaluated using two enzymatic systems, yielding 3 µm crystals of lysozyme and 2 µm crystals of Pdx1, an Arabidopsis enzyme involved in vitamin B6 biosynthesis. A seeding strategy was used to overcome the improbability of Pdx1 nucleation occurring with diminishing droplet volumes. Convection within droplets was exploited for rapid crystal mixing with ligands. Mixing times of <2 ms were achieved. Droplet microfluidics for crystal size engineering and rapid micromixing can be utilized to advance time-resolved serial crystallography.

Keywords: crystal miniaturization; droplet microfluidics; micromixing; time-resolved serial crystallography.

Figure 1

Lysozyme crystal size control by droplet volume scaling. (a) Protein crystallization droplets generated at kilohertz frequencies by combining streams of lysozyme, mother liquor and fluorinated oil. (b) Using different devices and flow rates (see Table S1 of the supporting information), monodisperse droplets (CV < 4%) can be produced with picolitre to femtolitre volumes. (c) Lysozyme crystals produced under batch conditions (control, blue) were on average 8 µm long. The length of lysozyme crystals produced in droplets (salmon) correlates with the droplet volume, with ∼3 µm-long crystals produced in the smallest 0.89 pl droplets. Crystal uniformity emerges when the average number of crystals per droplet (λ) is ≤0.1. (d) and (e) Visual comparison of lysozyme crystals prepared in batch (control) and extracted from 0.89 pl droplets by breaking the emulsion. (f) Droplet volume miniaturization is associated with reduced crystal density normalized to crystals per nanolitre (green) which correlates with increasing the surface area to volume (SA:V, grey) ratio. The batch crystal density value is denoted by the green dashed line. (g) Gains in droplet generation frequency scale with droplet volume reduction.

Figure 2

The effect of seeding under batch conditions compared with droplet conditions on Pdx1 crystal size. (a) Seeded batch Pdx1 crystallization involved a 1:1:1 mixture of Pdx1, seed (10⁵–10⁷ ml⁻¹) and mother liquor. The seed dilution affects crystal size (blue), but not crystal uniformity. (b) Pdx1 crystals were grown in droplets using a 2:1 mixture of seeds (10⁷ ml⁻¹) in mother liquor with Pdx1. Pdx1 crystals grown in 219 and 18 pl monodisperse droplets typically have single occupancy (scale bars 100 µm). (c) Droplet miniaturization over a 200-fold range was used to control Pdx1 crystal length from ∼20 to ∼2 µm (salmon), with crystal length being proportional to droplet volume. (c, inset) Linear scaling of the crystal length with droplet diameter. Droplet confinement enables crystal-size uniformity (CVs 7.4–15.7%). Pdx1 crystals prepared in batch (control, blue) are large with low uniformity (CV 24.4%).

Figure 3

Mixing lysozyme crystals in droplets. (a) High-speed microscopy frame of mixing during droplet generation and transport along the channel (droplets are colour-enhanced to aid observation of mixing). The mixing rates with and without crystals are the same. Analysis involved 12 droplets with crystals and 10 droplets without. (b) Dye and crystal mixing in droplets at a ratio of 0.3 with a droplet velocity of 300 mm s⁻¹. Droplets containing crystals are highlighted with cyan circles. The mixing rate increases as the volume fraction of dye increases, with the optimal ratio being 0.5 (300 mm s⁻¹ droplet velocity). The droplet pixel intensity CV is plotted as the mean ± SD for 15 droplets. (c) Dye and crystal mixing in droplets at the optimal 0.5 ratio with a droplet velocity of 300 mm s⁻¹. The droplet pixel intensity CV is plotted as the mean ± SD for 15 droplets. Increasing velocity increases convection (circulations within droplets) and shrinks droplet volumes to reduce diffusion paths, with both causing faster mixing.