Frank
Schwemmer
*a,
Clement E.
Blanchet
b,
Alessandro
Spilotros
b,
Dominique
Kosse
c,
Steffen
Zehnle
c,
Haydyn D. T.
Mertens
b,
Melissa A.
Graewert
b,
Manfred
Rössle
b,
Nils
Paust
ac,
Dmitri I.
Svergun
b,
Felix
von Stetten
ac,
Roland
Zengerle
acd and
Daniel
Mark
c
aLaboratory for MEMS Applications, IMTEK - Department of Microsystems Engineering, University of Freiburg, Georges-Koehler-Allee 103, 79110 Freiburg, Germany. E-mail: frank.schwemmer@imtek.de
bEuropean Molecular Biology Laboratory, Hamburg Outstation, Notkestrasse 85, Hamburg, 22603, Germany
cHahn-Schickard – Georges-Koehler-Allee 103, 79110 Freiburg, Germany
dBIOSS Centre for Biological Signalling Studies, University of Freiburg, 79110 Freiburg, Germany
First published on 23rd February 2016
We present a centrifugal microfluidic LabDisk for protein structure analysis via small-angle X-ray scattering (SAXS) on synchrotron beamlines. One LabDisk prepares 120 different measurement conditions, grouped into six dilution matrices. Each dilution matrix: (1) features automatic generation of 20 different measurement conditions from three input liquids and (2) requires only 2.5 μl of protein solution, which corresponds to a tenfold reduction in sample volume in comparison to the state of the art. Total hands on time for preparation of 120 different measurement conditions is less than 5 min. Read-out is performed on disk within the synchrotron beamline P12 at EMBL Hamburg (PETRA III, DESY). We demonstrate: (1) aliquoting of 40 nl aliquots for five different liquids typically used in SAXS and (2) confirm fluidic performance of aliquoting, merging, mixing and read-out from SAXS experiments (2.7–4.4% CV of protein concentration). We apply the LabDisk for SAXS for basic analysis methods, such as measurement of the radius of gyration, and advanced analysis methods, such as the ab initio calculation of 3D models. The suitability of the LabDisk for SAXS for protein structure analysis under different environmental conditions is demonstrated for glucose isomerase under varying protein and NaCl concentrations. We show that the apparent radius of gyration of the negatively charged glucose isomerase decreases with increasing protein concentration at low salt concentration. At high salt concentration the radius of gyration (Rg) does not change with protein concentrations. Such experiments can be performed by a non-expert, since the LabDisk for SAXS does not require attachment of tubings or pumps and can be filled with regular pipettes. The new platform has the potential to introduce routine high-throughput SAXS screening of protein structures with minimal input volumes to the regular operation of synchrotron beamlines.
Author | Type | Minimum input volume | Automated generation of dilutions | # of different measurement conditions containing protein | Volume of protein per measurement condition | Time per measurement |
---|---|---|---|---|---|---|
a Via manual loading 5 μl are possible. b Estimated minimal fill for syringe pumps and tubings. c Time for measurement of 9000 droplets. d Six different measurement conditions (five containing protein) for 15 μl input volume. e Due to dilution, some of the measurements are performed with lower protein concentration than the input stock solution. f Including the manual positioning of wells. Automated image recognition based positioning is estimated to reduce time per measurement to 3–5 s. | ||||||
Hura et al.6 | Sample changer (Berkley) | N/A | No | 1 | 12 μl | N/A |
Round et al.25 | Sample changer (EMBL) | 20 μl | No | 1 | 5–20 μl | 60 s |
David et al.5 | Sample changer (Soleil) | N/A | No | 1 | 6 μl | N/A |
Nielsen et al.7 | Sample changer (CHESS) | 10 μla | No | 1 | 10 μl | N/A |
Stehle et al.17 | Microfluidic system (droplet based) | >50 μlb | No | 1 | ∼9000 × 0.5 nl = 4.5 μl | 900 sc |
Lafleur et al.2 | Microfluidic system (Lab-on-a-Chip system) | 15 μl | Yes | 5d | 15 μl/5 = 3 μle | 30 s |
LabDisk for SAXS | Microfluidic system (centrifugal microfluidic) | 2.5 μl | Yes | 15 | 2.5 μl/15 = 170 nle | 30–60 sf |
Microfluidics offers the potential to provide high-throughput systems that reduce sample volumes and time per measured condition. Different microfluidic devices have been presented for SAXS. Continuous flow devices, where liquids are mixed and analyzed in micro-channels have been used for the measurement of folding and unfolding kinetics via time-resolved SAXS.9–12 In these methods the spatial distance after the micro-mixer corresponds to a time after mixing. By probing at different distances from the micro-mixer, kinetics could be resolved down to ∼100 microseconds,13 in comparison to multiple milliseconds in stopped-flow measurements.14 Furthermore, there are a number of customized microfluidic solutions for specific questions in SAXS, e.g. the formation of spider silk15 or the alignment of anisotropic particles in microchannels.16 Nevertheless, next to these specialized microfluidic solutions, some promising Lab-on-a-Chip devices have been presented that allow for high-throughput screening of protein structures under different environmental conditions with minimal protein volume.
Stehle et al. presented a droplet based microfluidic system for SAXS.17 The system demonstrated the formation and analysis of gold nanoparticles. The volume of a single droplet was only ∼0.5 nl and the data from 9000 droplets were averaged during one SAXS measurement. In the future the system can be integrated with well-studied droplet microfluidics operations, e.g. splitting, merging, micro-injection and mixing of droplets, to form more complex analysis systems. The bioXTAS chip, developed by Toft et al., allows for automated mixing of down to 36 μl of sample solution and read-out in a 200 nl X-ray chamber.18 This device was later substantially extended by Lafleur et al., who developed a Lab-on-a-Chip system for automated high-throughput sample preparation.2 The Lab-on-a-Chip system contained multiple rotary valves for the on-demand mixing of screening agents. Protein concentrations could be continuously verified using UV absorbance measurements and protein consumption was reduced by an on chip sample reservoir. This allowed a full analysis cycle with 6 different measurement conditions to be measured with a protein volume of only 15 μl.
However, while all microfluidic systems available offer unique advantages for SAXS, current systems suffer from typical “world-to-chip” interfacing problems, e.g. all systems require the manual attachment of tubes for syringe pumps. This makes the current Lab-on-a-Chip systems impractical for novice users and calls for more user-friendly solutions. Furthermore, with the exception of the new version of the bioXTAS chip, all microfluidic systems have large dead volumes of tens to hundreds of microliters.
Here, we present the centrifugal microfluidic LabDisk for SAXS. In contrast to pressure-driven microfluidic approaches, centrifugal microfluidic systems can be filled with regular pipettes and intrinsically benefit from low dead volumes since no liquid is lost in tubings.19 This simple “world-to-chip” interface and low dead volumes have led to the adoption of centrifugal microfluidic platforms in diverse diagnostic20,21 and analytical22 applications. Especially noteworthy are centrifugal microfluidic systems for protein crystallography.23,24 The LabDisk for SAXS is a new high-throughput tool for SAXS experiments at the P12 beamline at EMBL (PETRA III, DESY). The protein volume for a full dilution matrix of 20 measurement conditions (15 protein solutions, 5 buffers) is 2.5 μl. This volume corresponds to 170 nl per protein containing measurement condition, which is more than 10× less protein volume per measurement in comparison to current sample changers and other microfluidic systems (see Table 1). The LabDisk for SAXS includes six dilution matrices, each offers automatic and precise aliquoting, merging and mixing of different combinations from three input liquids: the protein sample (2.5 μl), a screening agent (3 μl) and a buffer solution (3.5 μl).
We demonstrate that the LabDisk for SAXS can be used with a range of different liquids relevant for SAXS. SAXS data collected using the LabDisk for SAXS can be fitted against crystallographic structure and data quality is sufficient to calculate ab initio 3-dimensional structures for glucose isomerase. Finally, we show a first application of the presented LabDisk for SAXS, via demonstration of different intermolecular effects for glucose isomerase depending on salt and protein concentration. The LabDisk for SAXS has the potential to introduce routine high-throughput screening to a wide SAXS community, reducing protein use and measurement time at synchrotron beamlines for all users.
Fig. 1 LabDisk for SAXS. Each of the six segments includes the aliquoting of the three input liquids, the combination and the mixing in different predefined concentrations. The mixtures then reside in the read-out chambers. Read-out can be performed on disk in a synchrotron beamline. The fluidic function of the segments is explained in Fig. 2–5. |
The LabDisk for SAXS is inserted into the processing device. Reagents are pipetted into the three inlets of each of the dilution matrices. One to six dilution matrices can be run in parallel. Each dilution matrix (see Fig. 2) is filled with 2.5 μl of protein solution, 3 μl of screening agent and 3.5 μl of buffer solution. After pipetting, the lid of the processing device is closed, and the processing device starts. The disk is initially rotated at 10 Hz and slowly accelerated to 30 Hz over a 90 s period. Liquid is transported from the inlet into the inlet channel. The liquid then fills individual aliquoting fingers that are connected via an isoradial feeding channel. The liquid flow from the aliquoting fingers into downstream fluidic elements is stopped by geometric valves26 (32 μm × 20 μm) located at the radially outer end of each of the aliquoting fingers (Fig. 3A). Each aliquoting finger has a volume of 40 nl. After the aliquoting fingers are filled, excess liquid flows over a radial extension of the feeding channel into the waste. The radial extension of the feeding channel increases the hydrostatic pressure from the inlet to the waste and ensures that liquid from the feeding channel is completely drained. When the feeding channel drains, the individual aliquoting fingers are fluidically separated, completing the metering of the aliquots (Fig. 3B). After the metering is completed, the rotational frequency is increased to 150 Hz. Centrifugal pressure increases until the liquid bursts through the geometric valves. Liquid aliquots from the upper aliquoting row (protein sample) and the center aliquoting row (screening agent) are transported over the backside foil and combined with aliquots from the lower aliquoting row (buffer) (Fig. 2). Each of the six dilution matrices on the disk generates 120 aliquots. Aliquots are combined in six aliquots each with different ratios of protein solution, screening agent and buffer aliquots, e.g. 3 protein solution aliquots, 1 screening agent aliquot and 2 buffer aliquots (see Fig. 4 for all combinations). The aliquots are mixed by reciprocation using a pneumatic chamber (see Fig. 5).27,28 The rotational frequency alternates between 10 Hz and 150 Hz, pumping the liquid between the read-out chamber and the pneumatic chamber until the liquid plugs are mixed completely. After 10 mixing cycles, the disk is stopped, the liquid is pumped into the read-out chambers and ready for analysis via SAXS. Each dilution matrix then contains 20 different measurement conditions. A full disk of six dilution matrices contains 120 different measurement conditions in 120 read-out chambers. Out of the 2.5 μl input volume 680 nl are analyzed in the actual SAXS measurement. The rest of the volume is included for pipetting tolerance, filling and draining of the feeding channel and as excess volume to ensure complete filling of the read-out chambers (see ESI† Table S1).
Fig. 2 One dilution matrix of the LabDisk for SAXS. The left side shows a top view of the dilution matrix, the right side shows a cross-section. Red circles indicate through holes in the center foil. After metering the three input liquids, the rotational frequency is increased. Aliquots from the sample and screening agent are transferred through holes in the center foil (red circles), transported over the backside fluidic layer and combined with the buffer aliquots in the read-out and mixing section. Aliquots are combined in 20 sets of six aliquots each, comprising different combinations of sample, screening and buffer aliquots. Arrows indicate liquid flow. Side view is not to scale. Aliquoting is described in Fig. 3. The mixing is described in Fig. 5. The different target concentrations, as generated in one dilution matrix, are shown in Fig. 4. |
Fig. 3 Aliquoting principle. During filling the rotational frequency is slowly ramped up from 10 Hz to 30 Hz. When liquid from the inlet fills the inlet channel, the aliquoting fingers having 40 nl volume each are sequentially filled via the feeding channel and excess liquid is transported to the waste (A). All extra liquid above the metering fingers is transported to the waste due to the radial extension of the feeding channel (B). By ramping up the rotational frequency to 150 Hz a certain number of the 40 nl aliquots is merged to create multiples of 40 nl volumes and transferred to the mixing and read-out chambers (C). The aliquoting structures for protein solution, screening agent and buffer aliquoting contain 33 aliquots, 36 aliquots and 51 aliquots, respectively. Combination of aliquots is explained in Fig. 2, mixing is explained in Fig. 5. |
After the fluidic protocol is completed, the disk can be transferred from the processing device to the positioner within the P12 beamline (PETRA III, DESY). Each dilution matrix contains 20 read-out chambers. The alignment of the read-out chambers within the SAXS beamline is performed via a custom built 3-axes motor stage. The 3-axes stage contains a rotational motor, which is used to roughly align the read-out chambers with the beam. Two linear motors, accurate to 20 μm, are used for fine alignment of the measurement chamber with the X-ray beam. The alignment was performed by manually controlling the motors for this manuscript, but will be automated in the future. The depth of the measured liquid column in the read-out chamber is 860 μm, the minimum diameter of the read-out chamber is 348 μm. The size of the X-ray beam was 50 μm × 50 μm. The alignment is supported by an in-axis camera setup. For identification each read-out chamber is individually numbered via a bitcode next to the read-out chamber. Data is automatically collected. After all 120 measurements are completed the disposable disk can be discarded and exchanged with the next disk.
LabDisks consist of three thermally bonded layers, a frontside foil containing most of the microfluidic structures, a backside foil connecting the radially inner aliquoting structures with the radially outer mixing chambers, and a center foil (160 μm, Topas COC 8007 and Topas COC 6013 co-extruded compound foil) with drilled holes connecting the frontside and backside fluidic layer. Microfluidic structures in the frontside and backside fluidic layers were designed using the computer-aided design software SolidWorks 2011 (Dassault Systèmes SOLIDWORKS Corp., France) and micro-milled using a KERN Evo (KERN Microtechnik GmbH, Germany) into a PMMA master (Plexiglas, Evonik, Germany). A negative of the microfluidic structures was replicated in PDMS (Elastosil RT-607, Wacker Chemie). The frontside and backside foils consist of custom-made 120 μm thick co-extruded Topas COC 8007 and Topas COC 6013 compound foils. Microstructures were replicated via thermoforming on a custom-built hot-press. The thickness of the X-ray viewing windows after thermoforming was ∼70 μm for the frontside and ∼115 μm for the backside foil. Frontside, center and backside foil were aligned and bonded by means of thermal bonding, using pressurized air in a hot press.
The X-ray beam (energy 10 keV) was collimated to an effective beam size of about 50 μm × 50 μm at sample position, yielding a flux of 5 × 1011 photons per second. Data were collected on a Pilatus 2 M detector (Dectris, Villingen), with a sample-to-detector distance of 3.1 m covering a range of momentum transfer 0.02 nm−1 < s < 4.8 nm−1 (s = 4πsinθ/λ, where 2θ is the scattering angle, and λ = 0.12 nm is the X-ray wavelength). On each read-out chamber, 20 frames of 50 ms exposure time were collected, radially averaged and normalized to the transmitted beam intensity. Individual frames were manually inspected and compared to identify radiation damage. Frames exhibiting significant differences in intensity were discarded.
The “batch-mode” measurements were performed using an in-vacuum flow through cell coupled to an autosampler robot.14,25 Twenty frames of 50 ms were collected and analyzed using the SAXS data analysis pipeline.32
For the SAXS experiments, only the protein and screening agent concentrations in the read-out chambers matter, the absolute volumes are irrelevant. To measure the actual variation of protein concentration in the read-out chamber, we quantified the relative protein concentrations in the read-out chamber from the SAXS scattering data for the dilution matrix with glucose isomerase. Relative protein concentrations were quantified by measuring the actual scaling factors vs. expected scaling factors in the SAXS scattering pattern. The variation of measured protein concentration at the three different diluted target concentrations was 2.7–4.4% CV. This variation includes variations in aliquoting, combination and mixing. Due to the geometric valves in the LabDisk for SAXS, liquids with very low advancing contact angles on the COC surface (<45°) cannot be used in the LabDisk for SAXS, i.e. liquids with high concentrations of surfactants. For such low advancing contact angles, there is cornerflow in accordance with the Concus–Finn condition.36,37 This leads to unpredictable filling of channels and in extreme cases even “bursting” of geometric valves before the rotation of the disk is started. Consequently, the aliquoting fails. Another limiting factor is viscosity. Increasing the viscosity of input liquids will reduce the flow rate in the cartridge. At some point, the liquid flow rate will be so low that aliquoting would not be finished before the rotational frequency is increased to 150 Hz. Liquid remaining in the inlet would then be transferred to the first read-out chamber. The liquid with the highest viscosity tested was water with 10% glycerol, with a viscosity of 1.31 mPa s at 20 °C. If liquids of higher viscosities need to be handled, the frequency protocol can be adapted to include longer holding times. In general, failure of the fluidics is clearly visible from uneven filling of the read-out chambers and lack of liquid in one or more of the three waste chambers.
Read out chamber | Protein concentration (mg ml−1) | NaCl concentration (mM) |
---|---|---|
1 | 11.0 | 0 |
2 | 0.0 | 250 |
3 | 1.8 | 333 |
4 | 3.7 | 250 |
5 | 5.5 | 83 |
6 | 0.0 | 83 |
7 | 1.8 | 0 |
8 | 3.7 | 167 |
9 | 5.5 | 250 |
10 | 1.8 | 83 |
11 | 3.7 | 0 |
12 | 0.0 | 333 |
13 | 5.5 | 0 |
14 | 0.0 | 167 |
15 | 1.8 | 250 |
16 | 3.7 | 83 |
17 | 5.5 | 167 |
18 | 1.8 | 167 |
19 | 0.0 | 0 |
20 | 3.7 | 333 |
SAXS data were collected for the 20 different conditions, investigating protein dilution and the effect of NaCl concentration on intermolecular interactions. The radially averaged curves collected on the different read-out chambers of the disk were consistent with each other and had the same background except for one chamber that exhibited a very intense signal at a low angle (<0.4 nm−1). The 2D images collected on this read-out chamber show a strong parasitic signal at low angle suggesting that the chamber was not properly aligned with the beam. This data set was not considered in the following. For the consistent read-out chambers, scattering data were collected from protein and buffer solutions and difference profiles generated for analysis. The scattering curves corresponding to the dilution matrix of GI without additional NaCl (where the differences between the curves are the largest) are depicted in Fig. 7A.
The decrease in signal-to-noise ratio can be explained by (i) an increase of the instrumental background (ii) a decrease of the SAXS signal.
(i) In the sample changer operation, the capillary in which the sample is loaded is in vacuum. The LabDisk for SAXS is operated in a 3 cm wide air gap. In addition to the scattering of air around the disk, two vacuum windows were added on the beamline to break the vacuum path (detector flight tube window: Kapton, 30 μm; on-axis camera window: polycarbonate, 125 μm). Although these windows have been chosen because of their low scattering property, they do contribute to the experimental background.
(ii) The total SAXS signal is lower because the path length of the X-ray in the capillary is reduced: the path length for the read-out chamber is 860 μm, whereas the path length for the sample changer capillary is 1.7 mm. For the photon energy used in these experiments (10 keV), the optimal path length would be 1.88 mm,38i.e. the path length is closer to optimal in the sample changer capillary than in the disk. Furthermore, for the presented experiments, the beam was cut down to 50 μm × 50 μm to have the full beam well centered in the read-out chamber, which reduced the intensity of the incoming beam by approximately a factor of 10.
Even though the signal-to-noise ratio is reduced, the resulting SAXS profiles collected with the LabDisk for SAXS can be readily used for advanced modeling methods such as envelope determination. Fig. 7C shows an ab initio bead model reconstructed from the LabDisk for SAXS data (5.5 mg ml−1, 250 mM NaCl) using a P222 symmetry overlaid with the atomic structure of the GI tetramer (PDB ID. 1OAD). The SAXS model and high-resolution structure overlap well, showing that despite the increased noise, the disk data can be used to determine an accurate solute envelope ab initio. A comparison of the predicted scattering profile from the atomic structure and the experimental data is shown in Fig. 7C, demonstrating the very good fit of the atomic structure of tetrameric GI to the LabDisk for SAXS data (χ2 = 0.75, see ESI† eqn (1)).
A clear decrease of the computed Rg is observed when the protein concentration increases: from 3.3 nm at 1.8 mg ml−1 to 2.4 nm at 11 mg ml−1. Such behavior is characteristic of systems showing strong intermolecular repulsion. This observed decrease in Rg with protein concentration is significantly reduced in the presence of salt. At a GI concentration of 5.5 mg ml−1 the apparent Rg increases from 3.1 to 3.3 nm as the salt concentration is increased to 250 mM (Fig. 8B), in agreement with the Rg observed for dilute GI (<5.5 mg ml−1) in the absence of salt. At lower protein concentrations no large variation of the Rg is seen upon salt addition, presumably due to the effective protein concentration being below the critical value required for significant intermolecular repulsion to be observed.
The change in the apparent Rg of GI tetramers determined from the LabDisk for SAXS data is clearly due to a partial ordering of the protein molecules in solution, characteristic of repulsive intermolecular interaction. Interactions between proteins in solution are readily seen by SAXS, notably in the low angle region (s < 0.1 nm−1), where the observed scattering profile deviates from that of the form factor (the curve of an ideal solution at infinite dilution). Repulsion between the solutes results in a decrease in the SAXS signal at small angle. We applied a Guinier approximation to detect changes at low momentum transfers. From the resulting change in Rg, interactions between the solutes can be deduced, e.g. a decrease in the apparent Rg corresponds to a stronger repulsion between the solutes.
As glucose isomerase has an overall negative surface charge at neutral pH (pI ∼ 3), the repulsive interaction observed here can be attributed to electrostatic repulsion between the charged GI tetramers in solution. These repulsions are modulated by the addition of salts, which screen the charges of the protein:
- When the protein concentration increases (i.e. distance between solutes decreases), the interaction between the solutes increases, resulting in the decrease of the apparent Rg.
- At intermediate protein concentration (5.5 mg ml−1), the repulsions are still present and affect the Rg. When the salt concentration increases, the charges of the solute are screened, there is less repulsion between the solutes and the apparent Rg increases.
- For a lower concentration, the distances between the proteins become too large to see the interaction and no clear effect of the salt is observed.
For the future, we plan to optimize the integration of the LabDisk for SAXS platform in the BioSAXS beamline P12. This includes automated positioning, assisted by image recognition, which will reduce the time per measurement to 3–5 s. Furthermore, we have fabricated 200 LabDisks for SAXS with a total of 1200 dilution matrices. These disks will be made available to regular users of the P12 beamline (PETRA III, DESY).
Footnote |
† Electronic supplementary information (ESI) available: Photographs of the disk in the beamline, the processing device, additional SAXS data, the formula for calculation of χ2 and information on dead volume in the disk. See DOI: 10.1039/c5lc01580d |
This journal is © The Royal Society of Chemistry 2016 |