Measuring the Nucleation Rate of Lysozyme using Microfluidics

From Soft-Matter
Jump to: navigation, search

S. Selimovic, Y. Jia, and S. Fraden

"Measuring the Nucleation Rate of Lysozyme using Microfluidics"

Crystal Growth & Design, DOI: 10.1021/cg800990k (2009).

Entry by Meredith Duffy, AP 225, Fall 2011

Keywords: nucleation, crystallization, microfluidics, supersaturation, osmosis


The crystallization process can be divided into two component processes, nucleation and crystal growth. Traditionally, decoupling these processes and studying them independently has been difficult to accomplish, but is important for understanding and optimizing crystal growth in proteins. In a previous paper, the Fraden lab introduced the PhaseChip, a high-throughput microfluidic device that decouples the two processes using osmotic gradients. First, the proteins are highly supersaturated by generating the diffusion of water out of the oil-suspended droplets in which they are contained, inducing the proteins to form small crystals. Then the level of supersaturation is lowered, and the small crystals serve as homogeneous nucleation sites around which the proteins form larger crystals. The PhaseChip's advantages over other systems for crystallization studies are its high throughput, the reversibility and non-thermal mechanism of its saturation control method, and the small quantities of protein it requires (nL vs μL scale).

Here, Selimovic et al. employ the PhaseChip to statistically study nucleation rates of lysozymes under a range of crystallization conditions and compare their results to both theory and experimental nucleation rates determined elsewhere using thermal methods of saturation control.

Methods and Results


The PhaseChip employs a two-level system of independently controlled microchannels separated by a PDMS membrane permeable only to water. Protein solution containing a small concentration of NaCl enters the upper "storage" channels and is broken into 900 pL - 2.8 nL droplets via the periodic injection of surfactant-containing oil through a cross-junction. The droplets, now separated by oil and stabilized by surfactant, find their way into 2.8 nL wells attached to the channel sides, one per well; there are about 62 wells per independently controlled microchannel. The lower "reservoir" channels contain preset concentrations of salt water; an osmotic gradient is created between the channel layers and water flows into or out of the droplets (depending on the gradient direction), changing the concentration of protein. The storage channels are connected to a continuous supply of oil to compensate for these changes in droplet volume.

To create different levels of protein supersaturation, the authors flowed a 4M NaCl solution through the reservoir channels of five upper channels, exchanging the solution for a more dilute solution, one channel at a time, after a set amount of time ("quench time") ranging from 4 hours for the leftmost channel to 6 hours for the rightmost channel. Because the diffusive equilibration process has an exponential decay time constant of about 30 h, the supersaturation in each channel was continuously varying. Calculating protein and salt concentration in each droplet from the optically-measured droplet area, the authors then determined supersaturation as the ratio C/C_s of protein concentration C to the protein solubility C_s for the given salt concentration in the droplet (taken from literature). The supersaturation curves for the five channels are depicted in Figure 1(b), and the qualitative differences between the five channels in protein crystal formation after 118 hours can be seen in Figure 2, where the black dots are the crystals.


In accordance with other researchers, the authors quantified the nucleation sites for each quench time by counting how many drops contained "m" large crystals and normalizing over the total number of drops. Although their attempt to fit this data to a Poisson distribution failed, they did nonetheless find as expected that the longer the quench time, i.e. the higher the protein supersaturation, the more crystals nucleated. The mean number of crystals per drop found long after quench is plotted as a function of quench time in Figure 5 (black) and compared to theory (red).


Finally, the authors fit their data to a classic nucleation model to determine nucleation rate J as a function of time for supersaturation curve, then as a function of supersaturation at C_s=5 (Figure 6). Their results, plotted in green, find J to be several orders of magnitude smaller than results published elsewhere in literature, plotted in red. Possible sources of the discrepancy include:

  • heterogeneous nucleation
  • protein sticking to the wall of the PDMS well when the droplet shrinks (because the droplet sticks to the wall) and thus lowering supersaturation in the droplet
  • assumptions and approximations made in fitting the data to the nucleation model
  • the continual variation in supersaturation due to the slow rate of osmosis (whereas the other study used temperature to rapidly control supersaturation), which makes it difficult not only to compare results, but also to know whether the proteins have been at high supersaturation long enough to form enduring nuclei that don't dissolve at lower supersaturation before crystallization occurs.


The PhaseChip's diffusion-based system of determining crystal nucleation rates for proteins provides the advantages of high throughput (62 wells per condition, 5 conditions at once), low volume, and ability to explore parts of the phase curve that thermal methods cannot. However, its choice of diffusion to control supersaturation also prevents rapid changes, adding a level of complexity and uncertainty to the system. Some potential improvements are obvious, such as employing a better surfactant to keep droplets from attaching to the well and thus ensuring that all protein stays inside when they shrink, but others, such as getting a better handle on how long each well was at nucleation-inducing supersaturation levels rather than just estimating, are potentially entire studies in themselves. Nonetheless, although the results are quantitatively questionable, the system certainly has some validity as a qualitative assessment of nucleation and crystal formation, and could thus serve useful (even without improvements) for comparisons of nucleation rate between different proteins or other similar studies.