Real-Time Measurement of Protein Crystal Growth Rates within the Microfluidic Device to Understand the Microspace Effect.

Preparation of high-quality protein crystals is a major challenge in protein crystallography. Natural convection is considered to be an uncontrollable factor of the crystallization process at the ground level as it disturbs the concentration gradient around the growing crystal, resulting in lower-quality crystals. A microfluidic environment expects an imitated microgravity environment because of the small Gr number. However, the mechanism of protein crystal growth in the microfluidic device was not elucidated due to limitations in measuring the crystal growth process within the device. Here, we demonstrate the real-time measurement of protein crystal growth rates within the microfluidic devices by laser confocal microscopy with differential interference contrast microscopy (LCM-DIM) at the nanometer scale. We confirmed the normal growth rates in the 20 and 30 μm-deep microfluidic device to be 42.2 and 536 nm/min, respectively. In addition, the growth rate of crystals in the 20 μm-deep microfluidic device was almost the same as that reported in microgravity conditions. This phenomenon may enable the development of more accessible alternatives to the microgravity environment of the International Space Station.

Introduction

The production of high-quality crystals of inorganic, organic, and biological macromolecular compounds is important for the development of functionalized materials and pharmaceuticals as well as for quality control in the food industry. Crystallization condition screening and precise control of the crystallization environment are critical factors in the production of high-quality crystals. Microfluidics offers a high-throughput approach to this screening with low sample consumption.[1−3] This has many advantages in the field of materials science, particularly in relation to quantum dots, pharmaceuticals, and protein crystallography.[4−6]

On the other hand, natural convection is an uncontrollable phenomenon that occurs during ground-based crystallization and is one of the major factors influencing the quality of crystals after the optimization of a crystallization condition. After a crystal grows to a certain size, a concentration gradient forms around the crystal because of the decrease in solute molecules in solution. Solute molecules are then transported by diffusion along the concentration gradient. From the viewpoint of crystal engineering, high-quality crystals with higher-ordered molecular packing and fewer lattice defects are produced under the controlled concentration environment. However, natural convection disturbs the concentration gradient, meaning that crystals cannot grow under the controlled crystallization environment. To reduce this effect, microgravity has been employed to create a diffusion-controlled crystallization environment; however, protein crystallization is challenging in microgravity-based conditions.[7−9] Although protein crystals that are formed under a microgravity environment have shown improvements in resolution, mosaicity, and other crystallographic parameters, a true microgravity environment only exists on the International Space Station. Therefore, the development of imitated microgravity conditions on the ground including appropriate magnetic and electric fields has provided a more accessible approach to obtaining high-quality protein crystals for the collection of precise three-dimensional (3D) structural data.[10−14]

Microfluidic devices are useful tools used for protein crystallography[15−17] due to the ease of high-throughput screening, crystallization control, and on-chip X-ray diffraction measurement.[1,2,18−25] We have previously reported the effect of a microspace on the growth of protein crystals using microfluidic devices of 10 and 50 μm in depth.[26] In the previous study, the use of a microfluidic device with a depth of 10 μm prevented cluster-like and needle-like crystal morphologies of glucokinase from Pseudoalteromonas sp. AS-131 (PsGK) and produced plate-like crystals of PsGK, which were suitable for X-ray diffraction analysis. We assumed that this confined crystallization space induced changes in crystal morphology by a diffusion-controlled crystallization environment similar to microgravity and reducing solute transportation.

Generally, microfluidic devices of 100 μm in depth can be regarded as the microgravity environment. Several research groups reported the microfluidic device for protein crystal growth in microgravity and the typical optical microscope-based crystal growth analysis.[27,28] However, the mechanism and kinetics of protein crystal growth in the microfluidic device was not elucidated in detail due to limitations in measuring the crystal growth process within the device. To overcome the problem, we demonstrated real-time spectroscopy-based measurement of the protein crystal growth in the microfluidic device. We conducted a detailed investigation of the growth of lysozyme crystals, as a typical model protein, using microwells of 10, 20, 30, and 50 μm in depth. To achieve this, we visualized the concentration gradient using in situ Raman spectroscopy[29] and determined the growth rate of the lysozyme crystals inside the microfluidic device using a noninvasive approach of laser confocal microscopy combined with differential interference contrast microscopy (LCM-DIM) on a nanometer scale.[30,31] From microscopy-based measurements, we confirmed the formation of a concentration gradient around the lysozyme crystal and found the difference of the normal growth rate of the crystal in the 20 and 30 μm crystallization space. The crystals from both conditions were analyzed by on-chip X-ray diffraction at a synchrotron facility, and the lysozyme crystal prepared in the 20 μm-deep microfluidic device provided a highly reliable electron density map.

Results and Discussion

Real-Time Visualization of Protein Crystal Growth in the Microwells

We predicted that the environment of the microspace would lead to reduction of the solute transportation caused by natural convection. The Gr number, which relates to natural convection, became small enough to suppress natural convection because the microwell depth (10 μm: 10–5 m) is the largest factor (eq )where g represents the gravitational acceleration, ß is the solution expansivity coefficient, Cs is the protein concentration of the surface, C is the protein concentration of the fluid, L is the depth of the microwell, and ν is the kinematic viscosity of the fluid. According to eq , the microfluidic devices of 100 μm in depth can be regarded as the microgravity environment. Therefore, molecular diffusion in the microwells is dominated by Fick’s law of diffusion.

We attempted to visualize the protein concentration profile in the 100 μm-deep microwell because the microwell size smaller than 100 μm might be enough to suppress natural convection. In addition, the nucleation frequency depends on the volume of the crystallization solution in the microspace. Therefore, we decided to use the 100 μm-deep microwell for the visualization experiment in this step. The protein concentration profile was visualized using in situ microscopic Raman spectroscopy.[29] The concentration profile in the microwell provides indispensable information relating to protein crystal growth in the microspace because supersaturation plays the key role in crystallization. Figure a presents a bright-field image and cross-sectional view of the measurement system. The lysozyme crystal exhibited two main Raman shifts at 1557 and 2940 cm–1, which correspond to the amide I band of the peptide backbone and the C–H stretching vibration of the peptide backbone, respectively. We selected a wavenumber at 2940 cm–1 as the measurement wavenumber, which showed a higher intensity compared with 1557 cm–1 (Figure S1). The base material of the window was also investigated to identify a protein concentration profile with a high signal-to-noise ratio. As shown in Figure S1, the polymer films that were used as substrates in the microfluidic devices exhibited a Raman shift at a wavenumber of around 3000 cm–1. Thus, we used a PDMS substrate sandwiched with glass substrates to prepare the device. The protein concentration profile was created from 81 points of Raman mapping data measured from 2900 to 3000 cm–1. The exposure time was 12 s, and the measurement region was 600 μm2 with 75 μm2 steps. The Raman spectrum of air was used as a baseline spectrum (Figure S2).

Figure 1

(a) Bright-field image and illustration of the setup for in situ microscopic Raman spectroscopy. Air, a lysozyme crystal, and lysozyme solution were analyzed simultaneously in the same field of view. (b) Time course of protein concentration profiles. High and low lysozyme concentrations are indicated with red and blue colors, respectively. The red area represents the presence of the lysozyme crystal, corresponding to the bright-field image. Definitions: PDMS, polydimethylsiloxane.

Figure b shows the time course of the concentration profile in the microwells. The lysozyme crystal showed higher Raman intensity than the lysozyme solution because the crystal comprises a high density of lysozyme molecules. We observed a lysozyme depletion region in the microwells at 4 h after incubation. The lysozyme concentration gradient after incubation remained constant for 4 h, in contrast to batchwise crystallization, due to suppression of natural convection. After 6 h, the concentration gradient was reduced by the molecular diffusion compared with that observed 4 h after incubation. The lysozyme concentration gradient around the crystal almost disappeared after 8 h of incubation (Figure S3). From these results, we provide qualitative confirmation of the presence of a lysozyme depletion region, namely, the low supersaturation region, around the growing lysozyme crystal. Therefore, low supersaturation is theoretically maintained in the microwells smaller than 100 μm in depth.

Lysozyme Crystal Growth

We evaluated the lysozyme crystal face that grew preferentially in the microwells. From the result of in situ Raman measurement, we used 10, 20, 30, and 50 μm-deep microwells (Figure ). These microwells were used for the formation of lysozyme concentration gradient as same as the 100 μm-deep microwells. We found the lysozyme crystal growth behavior to be influenced by supersaturation conditions; according to literature, supersaturation affects the crystal growth rate, preferential growth face, and orientation crystal surface.[32,33] A tetragonal lysozyme crystal exhibits two typical crystal facets, the {1 0 1} face and the {1 1 0} face, as shown in Figure (insets). The {1 0 1} face is preferentially grown in low supersaturation; thus, the {1 1 0} face was oriented parallel to the substrate. Figure a shows the percentage of crystal faces that were oriented parallel to the microfluidic device substrate and confirms that the depth of the crystallization wells affects the orientation of lysozyme crystals. In the case of the 10 μm-deep microwell, only the {1 1 0} orientation was seen. The percentage of this orientation decreased slightly (to 96%) when the microwell depth was increased to 20 μm. On the other hand, the percentage of the {1 1 0} orientation decreased to 60% in microwells that were 30 μm or more in depth. Although these microwells could be maintained at low supersaturation regardless of the microwell depth, the 10 and 20 μm-deep microwells showed the preferential orientation of the {1 1 0} face.

Figure 2

(a) Top-view of the microfluidic device with 48 microwells. (b) Three-dimensional perspective illustration of the microfluidic device. The magenta area in the bottom layer represents the fluid channel, and blue and green areas represent the vacuum channels of the control layer. (c) Cross-sectional view of the microfluidic device. We fabricated microfluidic devices with 10, 20, 30, and 50 μm-deep microwells. (d) Schematic illustration and photograph of the operating process. Lysozyme crystallization solution was pipetted onto the microfluidic device and introduced into the microwells simultaneously using a vacuum pump. Microwells and fluid channels were visualized using food dye. Definitions: COP, cyclic olefin polymer; h, height; PDMS, polydimethylsiloxane; w, width.

Figure 3

(a) Bar graph of the distribution of the lysozyme crystal face orientation in relation to the microfluidic device substrate. Blue and orange represent the percentages of (1 0 1)- and (1 1 0)-oriented faces, respectively. Inset: photographs of the (1 0 1) and (1 1 0) crystal face orientations. (b) Photographs of the time course of the growth of lysozyme crystals in 20 and 50 μm-deep microfluidic devices after incubation at 20 °C in an incubator for (left) 1 and (right) 24 h. Scale bars, 50 μm.

Figure b presents photographs of the time course of the tetragonal lysozyme crystals of the 20 and 50 μm-deep microwells. The growth axes of the {1 1 0} phase was similar for crystals grown in 20 and 30 μm-deep microwells. In these microwells, the X axes of the crystals were preferentially grown compared with the Y axes after 1 h of incubation. However, after 24 h of incubation, the preferential growth axis changed from the Y to the X axis in both the 20 and 30 μm-deep microwells. In the 50 μm-deep microwell, crystals grew without any change in growth axes. From these results, we can conclude that a difference of only 20 μm in microwell depth supersedes supersaturation, resulting in the observed protein crystal growth behavior. We also assume that the large surface-to-volume ratio might affect the protein crystal growth. Thus, the combined effect of the microspace and large surface-to-volume ratio has a critical role, which affects protein crystal growth even when crystallization conditions are constant.

Real-Time Measurement of Crystal Growth Kinetics in the Microwells

We attempted real-time measurement of lysozyme crystal growth kinetics to quantitatively evaluate the critical size of the microwell for protein crystal growth. The {1 1 0} face that was observed in the 20 and 30 μm-deep microwells was measured by LCM-DIM (Figure a).[31] The crystal growth rate in the height direction was calculated by counting the number of interference fringes as followswhere G represents the crystal growth rate, i is the number of interference fringes, λ is the wavelength of the light source, n is the refraction index of the crystal, and t is measurement time. We used a superluminescent diode with a 680 nm wavelength, and the refraction index of lysozyme was 1.49.[34] The displacement of interference fringes in both sizes of microwells were recorded, as shown in Movies S1 and S2 (20 and 30 μm, respectively). Figure b,c shows capture images from Movies S1 and S2. We observed one and seven interference fringes for the 20 and 30 μm-deep microwells at 323 and 182 s, respectively. The normal growth rates of {1 1 0} faces in the depth direction were calculated to be 42.2 and 536 nm/min for the 20 and 30 μm-deep microwells, respectively. The 10 μm difference in microwell depth induced an almost 10-fold change in crystal growth rate. Figure d depicts the relationship between lysozyme concentration and crystal growth rate of the {1 1 0} face in 3.5 and 5% NaCl solution under typical batchwise conditions, as described by Durbin and Feher, with slight modifications.[32] In the present study, the lysozyme crystallization solution contained 4% NaCl (0.7 M) as the precipitant and 4% (40 mg/mL) lysozyme solution. Therefore, we estimated the growth rate of the {1 1 0} face in conditions of 4% NaCl and 4% lysozyme solution (Figure d) to be 300–800 nm/min, which agrees with the experimental data from the 30 μm-deep microwell. However, the growth rate of the {1 1 0} face in the 20 μm-deep microwell was considerably different from the estimated value. From the protein concentration visualization experiment and the calculated Gr number, the 20 μm-deep microwell predicted to form the protein concentration gradient around the crystal. In addition, the growth rate measurements of the {1 1 0} face and the results of crystal orientations suggest that the crystallization in a space smaller than 20 μm is a unique situation compared with the other microspace environments.

Figure 4

(a) Schematic illustration of the measurement of the lysozyme growth rate. Lysozyme crystal growth was measured using laser confocal microscopy with differential interference contrast microscopy (LCM-DIM). (b, c) Capture images (from Movies S1 and S2) of lysozyme crystals grown in (b) 20 and (c) 30 μm-deep microfluidic devices captured using the LCM-DIM system. Interference fringes were observed depending on the crystal growth rate. One and seven fringes were observed for the 20 and 30 μm-deep microwells at 323.47 and 182.97 s, respectively. Scale bars, 20 μm. (d) Graph of the relationship between lysozyme concentration and lysozyme crystal growth rate of the (1 1 0) face. Closed triangles and closed squares represent the growth rates in 5 and 3.5% NaCl, respectively, as reported by Durbin and Feher.[32] The blue area indicates the estimated growth rate in conditions of 4% (0.7 M) NaCl and 4% (40 mg/mL) lysozyme. Blue and red broken lines indicate the lysozyme crystal growth rate in the 20 and 30 μm-deep microfluidic devices in the present study, respectively.

According to literature, the solubility of the lysozyme in the conditions used in the present study would be 3 mg/mL.[35] Thus, supersaturation is calculated to be 12 or 2.6 for (C – Ce)/Ce or ln(C/Ce), respectively (C represents the concentration of the lysozyme, and Ce represents the solubility of the lysozyme at that concentration). Suzuki et al. measured the lysozyme growth rate of the {1 1 0} face using laser interferometry and found the growth rate to be almost 96 nm/min at supersaturation ln(C/Ce) of 2.5.[36] The lysozyme growth rate we recorded in the 20 μm-deep microfluidic device was slightly slower than that reported for microgravity. The purity of the lysozyme affects the crystal growth rate.[37,38] The crystal growth rate of the high-purified lysozyme was larger than that of the low-purified lysozyme. The purities of lysozyme in this study and Suzuki et al.’s were >95 and 98.5%. For the purity effect, we consider that the crystal growth rate in the 20 μm-deep microfluidic device might be slower than that of the microgravity environment studied by Suzuki et al. This suggests that our microfluidic device enabled suppression of natural convection, creating a diffusion-controlled crystal growth environment.

We assume that the creation of controlled protein depletion at the crystal top-surface and reduction of solute transportation in the microspace are major factors for the reduction of lysozyme growth rate in the 20 μm-deep microfluidic device (Figure a,b). To the best of our knowledge, this is the first study to provide quantitively evidence that the crystal growth rate in a microfluidic device larger than 30 μm is similar to that observed in batchwise crystallization conditions. Although further investigations are required to elucidate the phenomenon, we consider the creation of controlled-protein depletion by the microspace and the large surface-to-volume ratio to have critical roles in the protein crystal growth. These findings may enable the development of novel imitated microgravity environments for crystal engineering in biology and life science.

Figure 5

Schematic illustration of the lysozyme crystal growth in (a) 30 and (b) 20 μm-deep microwells.

Figure shows the 3D structure of the lysozyme generated from data collected from the crystal grown in the 20 μm-deep microfluidic device, and Table shows a comparison of crystallographic statistics. The space group and unit cell of the lysozyme crystals in the present study are the same as those reported for lysozyme crystals prepared using typical batchwise methods.[39] To process X-ray diffraction data, we used a CC1/2 of around 50% as a standard value to determine the highest resolution shell.[40] The resolution limits were 1.43 and 1.36 Å for crystals grown in 20 and 30 μm-deep microfluidic devices, respectively. All other statistics were equivalent for the two crystal data sets, although the volumes were 1.5-fold different. It should be noted that the crystal size, including thickness, strongly affects the diffraction intensity and degradation of the crystal due to radiation damage. In particular, the diffraction intensity is proportional to the crystal volume. However, the lysozyme crystal formed in the 20 μm-deep device provided diffraction data, which was comparable with that collected from the lysozyme crystal grown in the 30 μm-deep device. Furthermore, the 20 μm lysozyme crystal showed a higher redundancy and lower Wilson B factor, which indicate temperature-dependent atomic vibration, than the 30 μm crystal. These results indicate the high reliability of the electron density map for the 20 μm crystal.

Figure 6

(a) Three-dimensional structure of the lysozyme generated from the crystal formed in the 20 μm-deep microfluidic device. (b) Three-dimensional lysozyme surface model and (c) enlarged surface model showing the active site, hydrated water molecules (red and white spheres for oxygen and hydrogen atoms, respectively), and hydrogen bonds (yellow broken lines). Blue and red colors represent positive and negative charges, respectively.

Table 1

Crystallographic Statistics of Lysozyme Crystals Grown in the 20- and 30-μM-Deep Devices

statistics	20 μm	30 μm
space group	P43212	P43212
unit cell (Å, °)	a = b = 79.2	a = b = 78.9
	c = 37.0	c = 37.0
	α = β = γ = 90°	α = β = γ = 90°
resolution (Å)	1.43–35.45 (1.31–1.39)	1.36–35.28 (1.36–1.45)
total reflections	275,014	288,292
unique reflections	41,197	47,391
redundancy	6.7	6.1
completeness (%)	99.9 (99.4)	99.6 (97.7)
I/σ (I)	4.2 (1.00)	11.6 (1.00)
CC1/2 (%)	98.7 (47.8)	99.8 (47.1)
Wilson B factor (Å2)	8.12	10.04
mosaicity (°)	0.136	0.104
refinement
Rwork (%)	0.220	0.199
Rfree (%)	0.237	0.232

Figure b,c shows the structures generated from the diffraction data. We confirmed that the negatively charged active site contained glutamic acid and aspartic acid. Water molecules were found to be densely bonded at the active site, forming hydrogen bonds. In total, 192 water molecules were assigned. The crystals prepared in the 20 μm-deep microfluidic device were easily collected by peeling off the top layer. Therefore, protein crystallization using this device coupled with the seeding technique enables preparation of large, high-quality protein crystals, and this combination has been reported in several papers.[20,26] The imitated microgravity environment of the microfluidic device will be useful like agarose gels and an efficient convection-free geometry to perform preliminary trials for real microgravity experiments on the International Space Station.[41,42]

Conclusions

In this study, we present a real-time analysis of protein crystal growth rates within the microfluidic devices. We conducted quantitative analysis of the difference in the crystal growth rate between 20 and 30 μm-deep microfluidic devices using microscopic methods and found the rates to be 42.2 and 536 nm/s, respectively. The growth rate of crystals in the 20 μm-deep microfluidic device was the same as that reported for the microgravity environment. Our observations of protein crystallization behavior and microscopic imaging as well as the statistics of diffraction data and 3D structure modeling suggest that a microspace smaller than 20 μm was a unique situation compared with the other microspace environments.

Microgravity environments have attracted attention in many scientific fields, although this is not available to many researchers. In particular, protein crystallization in microgravity is expected to result in improved crystal quality. However, these crystallization experiments have only been trialed at the International Space Station. We cannot predict the suitability of protein samples for microgravity-based crystallization experiments, which creates a problem due to the high cost and time-consuming process of the trials. Therefore, development of an imitated microgravity environment would offer a user-friendly microgravity environment for protein crystallography at the ground level. Our findings may accelerate the development of novel imitated microgravity without the need for any special apparatus or equipment.

Experimental Section

Materials

Hen egg-white lysozyme (>95% purity) was purchased from Hampton Research (Aliso Viejo, CA, USA). We used lysozyme without further purification. Sodium chloride, sodium acetate, acetic acid, glycerol, acetone, and 2-propanol were purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). Trichloro(1H,1H,2H,2H-perfuluorooctyl)silane was purchased from Sigma-Aldrich (St. Louis, MO, USA). Polydimethylsiloxane (PDMS; SILPOT 184 W/C) was purchased form Dow Corning Toray Co., Ltd. (Tokyo, Japan). We purchased a SU-83010, a SU-83050, and a SU-8 developer from Nippon Kayaku Co., Ltd. (Tokyo, Japan). Silicon wafers were obtained from Global Top Chemical (Tokyo, Japan).

Fabrication of Microfluidic Devices

Microfluidic devices with normally closed valves were fabricated by a standard soft lithographic procedure with minor modifications.[43] Silicon wafers were washed with acetone and 2-propanol prior to making SU-8 molds. We poured SU-8 onto the silicon wafers, controlling the thicknesses of the SU-8 layers using a spin coater (MS-A100, Mikasa Shoji, Tokyo, Japan) to obtain 10, 20, 30, and 50 μm SU-8 layers for the fluid layers (FL). The thickness of the control layer (CL) was 50 μm. The SU-8-coated silicon wafers were baked onto a hot plate, and then, photomasks (12,700 dpi, Unno Giken Co., Ltd., Tokyo, Japan) were aligned onto the wafers and exposed to ultraviolet (UV) light using a mask aligner (M-1S, Mikasa Shoji) to cross-link the SU-8. A non-cross-linked SU-8 was then developed with the SU-8 developer, and the molds were treated with trichloro(1H,1H,2H,2H-perfuluorooctyl)silane vapor. Polydimethylsiloxane was poured onto the SU-8 molds and spin-coated to obtain 70 μm-thick PDMS layers, which were then baked at 80 °C. The PDMS-coated SU-8 mold for the CL and a 40 μm-thick cyclic olefin polymer film (COP; Zeon Corporation, Tokyo, Japan) were treated with oxygen plasma (CUTE-1MPR, Femto Science, Gwangju, Korea), and then, the COP film was bonded to the PDMS layer on the mold. The COP-PDMS layer was cut out from the mold and aligned to the PDMS layer of the SU-8 mold for the FL. The SU-8 mold for the FL was baked at 80 °C. The COP-PDMS(CL)-PDMS(FL) layer was cut out from the mold and the bottom PDMS layer covered with a 40 μm-thick COP film to create microchannel structures. A photograph, 3D perspective view, and cross-sectional view of the microfluidic device are shown in Figure a–c.

Protein Crystallization

Lysozyme solution was prepared by dissolving an appropriate amount of lysozyme into 100 mM acetate buffer at pH 4.5. The protein concentration was measured using a NanoDrop instrument (ND-ONE-W, Thermo Scientific, Tokyo, Japan) and adjusted to 80 mg/mL. A precipitant solution of 1.4 M sodium chloride in 100 mM acetate buffer (pH 4.5) was prepared. Lysozyme and precipitant solutions were filtered through 0.2 μm syringe filters (Minisart RC4 or RC25, Sartorius Stedim Biotech, Gottingen, Germany). Equal volumes of both solutions were mixed to prepare the crystallization solution, which was then pipetted onto the inlet of the microfluidic device (Figure d). The crystallization solution was introduced into the microfluidic device using a vacuum pump; after loading into the microwells, inlets of the fluidic and control channels were sealed with a Crystal Clear sealing tape (HR-3-511, Hampton Research, Aliso Viejo, CA, USA). Microfluidic devices were incubated at 20 °C in an incubator (MIR-154-PJ, Panasonic, Osaka, Japan). Any protein crystals that formed in the microwells were observed on an optical microscope (ECLIPSE Ti-U, Nikon, Tokyo, Japan).

In Situ Microscopic Measurements

The concentration profile in the microwell was measured by microscopic Raman spectroscopy (XploRA, Horiba Co., Ltd., Kyoto, Japan). Raman spectra were collected from samples after 12 s of irradiation (at 532 nm). The measurement wavenumbers were 2800–3000 cm–1, and the detection area per measurement period was 600 μm2. Measurements were carried out in a temperature-controlled room at 20 °C. Although the supersaturation might shift slightly due to the temperature control system, we did not observe any change of crystals during the measurement. The lysozyme concentration profile was obtained based on a Raman intensity at 2900 cm–1, which corresponds to the C–H stretching vibration of the peptide backbone.[29] The protein crystal growth rate was measured using laser confocal microscopy with differential interference microscopy (LCM-DIM).[30,31,37,44] We measured the displacement of the interference fringes on the (1 1 0) face of lysozyme crystals formed in 20 and 30 μm-deep microwells to calculate the crystal growth rate. Protein crystals were prepared in the same manner, as described above. After several hours of incubation, we observed the microfluidic device using the optical microscope to find the measurable crystal by Raman spectroscopy and LCM-DIM. Then, in situ microscopic measurements were carried out immediately.

On-Device X-Ray Diffraction and Crystal Structure Analysis

On-device X-ray diffraction analysis has been reported previously; we carried out this analysis on beamline BL17A at the Photon Factory (Tsukuba, Japan) with minor modifications.[19] Lysozyme crystals were measured at a wavelength of 1.5 Å with a 1 s exposure time and a 1° oscillation step at 100 K with cryoprotection. We collected 180 images of X-ray diffraction data in which data were analyzed using XDS software and the CCP4 suite of programs.[45,46] Refinement of diffraction data was carried out using Coot based on the Protein Data Bank model 193L.[47] Structural models of the lysozyme were produced using PyMOL (the PyMOL Molecular Graphics System, version 2.1., Schrödinger LLC, NY, USA). We used a CC1/2 value of around 0.5 as the highest resolution limit of the diffraction data.[40]