Dielectrophoretic sorting of membrane protein nanocrystals

Abstract

Structure elucidation of large membrane protein complexes is still a considerable challenge, yet is a key factor in drug development and disease combat. Femtosecond nanocrystallography is an emerging technique with which structural information of membrane proteins is obtained without the need to grow large crystals, thus overcoming the experimental riddle faced in traditional crystallography methods. Here, we demonstrate for the first time a microfluidic device capable of sorting membrane protein crystals based on size using dielectrophoresis. We demonstrate the excellent sorting power of this new approach with numerical simulations of selected submicrometer beads in excellent agreement with experimental observations. Crystals from batch crystallization broths of the huge membrane protein complex photosystem I were sorted without further treatment, resulting in a high degree of monodispersity and crystallinity in the ~100 nm size range. Microfluidic integration, continuous sorting, and nanometer-sized crystal fractions make this method ideal for direct coupling to femtosecond nanocrystallography.

Figure 1

a) Schematic of the entire sorting device (without reservoirs for clarity). A single 100 μm inlet (I) channel is connected to five outlet channels (2 outer channels (O), 2 mid-outer channels (MO), 1 center channel (C)) where sorted fractions are collected. Positive potential (+HV) is applied to the inlet and negative potentials (−HV) are applied to outlets. The total device length is 5 mm. Scale bar is 1 mm. b) Zoomed in schematic of the constriction region connecting the inlet channel to the outlets. The 100 μm wide inlet converges into 30 μm evoking iDEP. In c), areas of high ∇E² are shaded, in which the largest DEP response is realized. Due to negative DEP, particles are repelled from these areas proportional to their DEP mobilities. Larger particles focus inward towards the center of the device, as shown by the thicker, solid arrows. Conversely, smaller particles that experience less F_DEP are deflected into the side outlet channels as illustrated with the thinner, dashed arrows. Scale bar is 20 μm.

Figure 2

Concentration distributions as obtained from numerical simulations (for details see the experimental section) of 90 nm and 0.9 μm particles in the microsorter at various potential schemes (+10V inlet in all cases). a) −20V in all outlets shows equal distribution for both particle sizes. b) −60V in the center outlet (−20V in all other outlets) with DEP shows focusing of 0.9 μm particles whereas 90 nm particles completely deflect. c) −60V in center outlet without DEP shows deflection of both particle sizes. b) and c) indicate the importance of DEP in the sorting mechanism. d) Increasing the potential in the center outlet to highly negative values (below −80V) can focus both particle sizes, indicating the importance of an optimal potential scheme. The color legend represents the concentration normalized to the inlet concentration. Scale bar is 100 μm.

Figure 3

a) Fluorescence microscopy snapshot showing the 90 nm beads distributed in all outlet channels when −60V is applied to the center outlet (−20V to all other outlets). b) Fluorescence microscopy snapshot of the 0.9 μm beads focusing at the same potential scheme as in (a). Scale bar is 50 μm. c) Quantified particle distributions in each outlet channel for both particle sizes as measured by fluorescence intensity for the 90 nm beads and particle counting of 0.9 μm beads (see experimental section for details). A relatively equal distribution is seen for 90 nm beads whereas 90% of the 0.9 μm beads focus into the center outlet. Error bars represent the standard deviation.

Figure 4

a) Fluorescence image of PSI crystal sorting. Large crystals are shown focusing in the center of the device and smaller particles (as indicated by bulk fluorescence) are deflected into side outlet channels. Scale bar is 50 μm. b) DLS heat map of the bulk crystal solution injected into the inlet and c) of the center outlet focused solution. In (b) and (c), a broad size distribution is determined ranging from approximately 80 nm to 20 μm. d) DLS heat map of the solution deflected into O and MO side outlets from the same experiment showing a narrower size distribution of fractionated nanocrystals around 100 nm in size.

Figure 5

a) Fluorescence image of the inlet reservoir and b) of the center outlet reservoir solution after sorting a highly polydispersed, larger volume sample (+60V inlet, −60V center outlet, −5V MO and O side outlets). Scale bars are 50 μm. In c) and d) a histogram of the size distribution from an imaging threshold analysis is shown in which a wide range of particle sizes from 800 nm to 20 μm are detected for both the bulk and center outlet solutions. The lower limit of detection for this method is 800 nm, therefore, nanocrystals below 800 nm could not be individually resolved.

Figure 6

a) Fluorescence microscopy image of the solution in the O outlet reservoir containing the deflected solution (same experiment as Figure 5). As observed, very few particles can be individually resolved compared to the bulk and center outlet reservoirs shown in Figure 5, indicating a high content of nanocrystals. Scale bar is 50 μm. b) SONICC image of the high volume sample indicating crystallinity of the sample after having passed through the sorting device, as indicated by the second harmonic generation signal observed. c) DLS heat map of the deflected solution mainly containing nanocrystals (~ 60–300 nm) with a small contribution from microcrystals. d) Histogram of the DLS measurement: The major peak represents crystals with radii of 100 ± 30 nm, and an overall distribution shows a radii range of ~ 60–300 nm. A small contribution by microcrystals of ~ 1 μm in size is also seen here.