As we all know, supersaturation level drives crystallization (McPherson and Gavira 2014). One of the most widely used methods to build a supersaturated state is through vapor diffusion (McPherson 2017). A small drop composed of a mixture of protein sample and precipitant reagent is placed in a vapor diffusion device with a large volume of precipitant solution as the reservoir. Typically, after mixing with protein sample, the small drop contains a lower precipitant concentration than the reservoir. To achieve equilibrium, water molecules vapor leaves the drop and eventually ends up in the reservoir solution. As water molecules leave the drop, the protein sample undergoes an increase in relative supersaturation. Both the protein and precipitant increase in concentration as water molecule leaves the drop to the reservoir. Equilibration will be finally reached when the precipitant concentration in the drop is approximately the same as that in the reservoir (Luft and Detitta 1997).

The process of equilibration also causes different supersaturated states in the solution. As a rule, it can be generalized as a crystallization phase diagram, which is also known as the Ostwald-Miers diagram (McPherson and Gavira 2014; Moreno 2017). Protein concentration and precipitant concentration are two critical parameters in the phase diagram of crystallization. These two variates divide the phase diagram into four primary regions based on the increments of supersaturation level from low to high, including the stable zone, metastable zone, labile zone, and precipitation zone (Vekilov 2012). In addition, with high precipitating agent concentrations, some spherulites (microcrystalline aggregates) may be observed at the upper labile zone border, which is thought of as the spherulites zone (Tanaka and Nishi 1989). Occasionally, oil droplets (phase separation) are observed in solutions with very high concentrations of precipitating agent or protein. This state occurs in the upper right corner of the phase diagram. It is known as a phase separation zone (Tanaka and Nishi 1989).

In general, crystallization is composed of two processes, nucleation and crystal growth (Kashchiev et al. 2005; Sauter et al. 2015a). Nucleation usually requires a higher supersaturated level than that necessary to sustain crystal growth. Only when a crystal nucleus is formed in the labile region, the remaining protein molecule can bind layer by layer to this growing center, which results in development of a mature crystal. With the protein consuming during nucleation, the supersaturation position experiences a path through labile region to metastable region. Here, the supersaturation level is just enough for crystal growth, and not inadequate to support nucleation. However, if the initial protein and precipitant concentration are too high, the solution supersaturation level reaches to precipitation region directly, and masses of protein molecules have no choice but to bow to the over-supersaturated pressure and aggregate as irregular precipitation. While sometimes the initial supersaturation pressure is not high, the vapor diffusion pressure is too strong, which often results in the crystal drop concentrated excessively fast. As a consequence, the nucleation rate turns out to be too fast or the protein directly precipitates out. Inversely, it is not a rare probability event in the process of crystallization where lowering the protein concentration only results in smaller crystals, but no bigger crystals when raising protein concentration. For another aspect, downregulating the precipitant concentration just shows the clear drops, inversely, upregulating precipitant concentration tends to appear as microcrystal or precipitation. Anyway, such kind of problems always be a bottleneck during crystal optimization. Actually, the crystal growth experiences a unique path when the solution supersaturation state goes through a series of regions on the phase diagram (McPherson and Cudney 2014). However, part of proteins only possess a narrowed metastable region but a wide labile region, in which the potential growing to large crystals is very limited (Dale et al. 2003).

Protein or protein fractions whose molecular weight ranges from 10 kDa to 100 kDa, for example, ligand binding domain or DNA binding domain, always has been a research hotspot in biological macromolecule crystallography. Pleckstrin homology (PH) domain of 3-phosphoinositide-dependent protein kinase 1 (PDK1) protein is one such representative protein, which includes 148 amino acids and its molecule weight is about 17.5 kDa (Garcia-Viloca et al. 2022). However, the crystal of PDK1 unlike other classic protein crystals, such as lysozyme, concanavalin A, and proteinase K, which are easy to harvest crystals in a lot of different conditions, and are usually used as experimental models in many crystallization research (Liu et al. 2011). During our optimization process, PDK1 PH domain is very liable to be over-nucleated, even if only suffers from a small degree of precipitant concentration increase, and results in numerous insufficient-size crystals. Twinned crystals are also observed in high protein concentrations. Moreover, PDK1 PH domain does not appear crystal, or only forms crumbly crystals with cracks in low precipitant levels. Above situations hint that PDK1 PH domain has a wide labile region and a narrow metastable region. That is the reason why we choose PDK1 PH domain as a research model in current study.

Adding additional water could be a potential method in crystal optimization, especially for the stubborn protein that only possesses limited crystallization space and route optimization methods cannot obtain diffraction quality crystal. This method does not alter crystallization conditions and does not require any sophisticated techniques, it is a very simple and easy-to-perform method. In this study, we examined whether this method affects protein crystal quality through a serial approach. At the same time, the underlying crystallization mechanisms are comprehensively reviewed, and the strategies that how to utilize the additional deionized water to optimize crystals also be presented in detail.

Materials

Restriction endonucleases and T4 DNA Ligase were obtained from New England Biolabs (Beverly, MA, USA). pET-42a (+) vectors was purchased from Novagen (Madison, WI, USA). Cloned Pfu DNA Polymerase was purchased from Stratagen (La Jolla, CA, USA). Primers used for PCR reaction and DNA sequencing of constructed plasmid were obtained from Cosmogenetech (Seoul, Korea). The Kits for plasmid extraction and DNA purification were purchased from Cosmogenetech (Seoul, Korea). For the purification of protein, GSTPrep FF 16-10 column, and HiPrep^TM 16/60 Sephacryl^® S-100 HR gel filtration column from GE Healthcare Life Sciences were purchased. All Materials of reagent were biotechnological grade.

Protein expression and purification

Human PDK1 PH domain (residues 411-556) fragment was cloned into pET-42a (+) vector between the restriction enzyme site SacII and KpnI. GST-tag was added at the N-terminus of PDK1 linked by a PreScission protease cleavage site (LEVLFQ/GP). The constructed plasmid was transferred into Escherichia coli BL21 (DE3) competent cell (Subedi and Park 2023). Plate the transformants onto an agar plate containing 50 mg/ml kanamycin antibiotic and incubate at 37°C overnight. After that, a single colony was picked up from the agar plate to Luria-broth (LB) medium containing 50 mg/ml kanamycin. Next, the E. coli cells were grown in a shaking incubator at 37°C until the OD₆₀₀ reached 0.5, then moved to a 25°C incubator for another half hour to cool down the temperature. Isopropyl-β-d-thiogalactoside (IPTG) in a 250 mM concentration was used to induce when the OD₆₀₀ came to 0.6. The E. coli cells were harvested after overnight incubation by centrifugation at 8,000 rpm for 20 min. The cell pellet was re-suspended in 200 ml lysis buffer (50 mM Tris-HCl pH 8.0, 1 mM DTT, 500 mM NaCl, 10% (v/v) Glycerol) and followed by sonication. The lysate was centrifuged at 15,000 rpm for one hour to remove the cell debris. GSTPrep FF 16-10 column was used for lysate binding on GE AKTA Start Chromatography System. Then, the resin was washed with 200 ml buffer A (50 mM Tris-HCl pH 8.0, 1 mM DTT, and 500 mM NaCl), and eluted by buffer B (50 mM Tris-HCl pH8.0, 10 mM reduced L-Glutathione and 500 mM NaCl). PreScission protease was used to cleave the GST-tag by incubating it together at 4°C overnight. Cleaved PDK1 was concentrated and loaded onto a HiPrep^TM 16/60 Sephacryl^® S-100 HR gel filtration column that equilibrated with size exclusion buffer (50 mM Tris-HCl pH 7.0, 1 mM DTT, and 300 mM NaCl) at a rate of 1 ml/min. The eluted fractionation from the UV peak was collected together and concentrated to 28 mg/ml by a 10 kDa cut-off centrifugal concentrator (Amicon Ultra-15, Millipore, USA). The purity of PDK1 PH domain checked by SDS-PAGE showed that more than 99% (Bayascas et al. 2008).

Crystallization

Using 0.2 M lithium sulfate, 30% PEG4000, and 0.1 M Tris hydrochloride (pH 8.5) as a starting condition for PDK1. The crystallization was performed at a 96-well sitting drop plate, by mixing 1 μl protein and 1 μl crystallization reagent on the drop well with another 100 μl crystallization reagent solution in the reservoir well. Totally 12 groups were set up, and each group repeated 4 wells. The first group was set as a control group that did not add any deionized water. Next, a serial of volume deionized water was added into the drops of the remaining 11 groups in an incremental way using 0.2 μl as an interval. Except for the volume of protein and crystallization reagent, the increased volume in the drops ranged from 0.2 to 2.2 μl. The sealed plate was placed in a 20°C incubator to allow it to gradually reach vapor-diffuse equilibration. Checking the plate under a microscope every day. When crystals appear, record their size and crystal number until they grow up to the maximum.

A higher supersaturated level is conducive to protein molecules being compressed more tightly in the crystal lattice producing better diffraction quality. However, a higher supersaturated level often leads to a more serious over-nucleate phenomenon. To explore whether adding additional water can improve crystal quality in over-nucleate conditions or not, another crystallization experiment was carried out. Based on the starting reservoir condition of PDK1 PH domain, other reagents’ concentrations were fixed but only the concentration of PEG4000 was gradually increased to simulate an over-nucleate situation. When the PEG4000 concentration reached 34% (w/v), already no observed crystal appeared or the drop only showed micro-crystal or amorphous aggregate. On this account, a new round of crystallization used the over-nucleate condition (0.2 M lithium sulfate, 34% PEG4000, and 0.1 M Tris hydrochloride (pH 8.5)) as a starting point. Crystal drop was also prepared by mixing an equal volume of protein and reservoir in a sitting drop plate. Adding additional deionized water into each well by an incremental method. Each increment was 0.2 μl, and each group repeated 4 wells. The sitting drop plate was also placed in a 20°C incubator and recorded every day until the crystal approached the maximum size.

All experiments used distilled water are formulated in Type 1+ ultrapure water (18.2 megaohm-cm resistivity at 25°C) and sterile filtered through 0.22 μm filters. Crystallization reagents and 96-well sitting drop plates come from Hampton Research. Crystal picture is captured by Nikon SMZ800N microscope and Leopard image analysis software.

Crystallization by adjusting drop volume

The experiment result shows that adding a small amount of water ranging from 0.2 μl to 0.8 μl (increment 10%-40%) to the protein-reservoir mix drop (2 μl) does not significantly decrease nucleation rate and increase crystal size (Fig. 1). Until the additional water volume come up to 1μl (increment 50%), the over-nucleate phenomenon appears to be effectively controlled. In comparison with the crystals that appeared in the initial condition (0.2 M lithium sulfate, 30% [w/v] PEG4000, and 0.1 M Tris hydrochloride [pH 8.5]), the crystals that produced by adding additional deionized water showed less nucleation and larger crystal size. More drops showed single crystals without adhesion to the adjacent crystals that much convenient for dealing with the crystal picking work. Moreover, the crystal surface defect decreased and became smoother. Observing under the microscope, the crystals produced by adding additional deionized water show fewer cracks inside and outside compared to the control group. In addition, thin plate-shaped crystals converted to three-dimensional tetragonal crystals with enough thickness for x-ray diffraction. The crystal rigidity increased and it can tolerate more mechanical strength when picked up by a loop.

Figure 1. Crystals are produced by adding different volumes of additional water. (A, B) No additional water; (C, D) water volumes range from 0.2 μl to 0.8 μl (increment 10%-40%); (E, F) water volume comes up to 1 μl (increment 50%); (G, H) water volume reaches to 1.2 μl (increment 60%); (I, J) water volume over 1.4 μl (increment over 70%). The scale bar on the bottom left-hand corner of each image represents 200 μm.

Then, further increasing water volume, the nucleate rate became lower. When the additional water volume reached 1.2 μl (increment 60%), only one or two crystals were produced in each drop (Fig. 1G, 1H). However the crystal size became very large and the width was usually more than 1,000 μm. As a mass of protein molecules binds to a limited amount of growth center layer after layer, it markedly raises the possibility of molecule misaligned stack, which inevitably causes twinning crystals. In our trials, some of the twinned crystals even can be distinguished by the naked eye under a light microscope. In some other cases, the texture of large crystals becomes loose due to the protein molecules’ out-of-order staking.

When we tried to continue to add more additional water, the volume was over 1.4 μl (increment over 70%) in the following groups. In consequence, the nucleation was completely repressed. The setup drops did not show any crystals and maintained clear even more than three months (Fig. 1I, 1J). It is worth mentioning that the other drops, more or less, all showed crystals within one week after setting up in the vapor diffusion plate. Due to adding too much water into the drop, the precipitant concentration is hard to reach supersaturated, and the drop stays in the stable zone even set up for a long time. If we provide an infinite volume of reservoir solution against the drop, and wait a long enough time, maybe we can see the crystal appear (Mikol et al. 1990). However, it is uneconomical for the crystallization practice.

The crystals produced from initial condition (0.2 M lithium sulfate, 30% [w/v] PEG4000, and 0.1 M Tris hydrochloride [pH 8.5]) are too small to be picked up for X-ray diffraction. The optimized crystals with larger size and sufficient thickness to resist mechanical disturbance and therefore those crystals were collected by loop and frozen in the liquid nitrogen, then performed the X-ray measurement. The diffraction results showed that the crystals optimized by adding additional water can exhibit enough good resolution (less than 2.5 Å). However, adding different volumes of additional water does not directly relate to the final diffraction quality. In some cases, large crystals could only show mediocre diffraction quality while the small crystals, if they can be picked up, could exhibit very good diffraction mode with excellent resolution. Maybe the diffraction quality is influenced by many factors, for example, the cryoprotectant selection, data collection strategy, manual operation error, and so forth. We cannot renew hastily that the large crystal is equal to good diffraction data, or that the large crystal is better than the small crystal. Whereas, the method shown here may provide a way out of microcrystal or deficient crystal optimization.

Crystallization at a higher supersaturated level

As for the over-nucleated condition (0.2 M lithium sulfate, 34% [w/v] PEG4000, and 0.1 M Tris hydrochloride [pH 8.5]), adding additional deionized water was also able to effectively improve the crystallization result. Albeit in the beginning stage adding a small volume of deionized water, the over-nucleated phenomenon has no significant improvement. The micro-crystals gradually become larger single crystals with the additional water volume increasing, and the crystal number also obviously decreased. When the additional water volume goes up to 1.6 μl (increment 80%), ideal crystals was harvested (Fig. 2). The crystals became clearly visible with satisfactory size, and the crystal shape was changed from rhombohedral to fusiform. We also got desirable X-ray diffraction results with the resolution of 1.8Å from those crystals. Nevertheless, when adding too much water that is over 2 μl (increment 100%), the drops maintain a clear state even after a long time incubation, which is similar to previous results.

Figure 2. The optimization of over-nucleate crystal condition. (A) Over-nucleated condition (0.2 M lithium sulfate, 34% [w/v] PEG4000 and 0.1 M Tris hydrochloride [pH 8.5]); (B) over-nucleated condition (0.2 M lithium sulfate, 34% [w/v] PEG4000, and 0.1 M Tris hydrochloride [pH 8.5]) with additional 1.6 μl water (increment 80%). The scale bar on the bottom left-hand corner of each image represents 200 μm.

Through adding additional water, the drop supersaturation level retreats to the metastable zone or stable zone where the nucleation cannot happen. As the water evaporates under the pressure of vapor diffusion, the droplets gradually condense. In the early equilibrium stage, the evaporation rate at the edge of drop is higher than that on the drop center, which usually causes a relatively higher supersaturated level around the drop edge. For this reason, nucleation first occurs at the edge of drop due to the “fringe effect”, which was found in crystallization of PDK1 PH domain (Fig. 3).

Figure 3. PDK1 protein crystal first appears at the edge of drop. The scale bars represent 200 μm in the main panel and 20 μm in the enlargement.

When the water molecules leave the drop through a vapor phase and ultimately approach to reservoir solution, the nonvolatile protein sample and precipitant reagent that is left behind in the drop have undergone a distinct increase in concentration (McPherson and Cudney B 2014). In a general case, the drop whose initial ratio is half-and-half will shrink to roughly half of its original volume after equilibrium, the protein concentration and precipitant concentration also double. Crystal growers can change the initial drop ratio to change the volume relationship between the protein sample and precipitant reagent in the drop, for example, increasing or decreasing the ratio of protein or precipitant (Luft et al. 2007). The final precipitant reagent concentration in the drop is going to be the same as the reservoir because the vapor diffusion system will seek a balance between the drop and the reservoir (McPherson and Cudney 2014). However, by this way, it not only alters the nucleation starting point on the phase diagram but also affects the final equilibrium state, which could bring the crystallization to another uncontrollable path. As we know, crystal drop final pH value depends on the reagent ratio of protein buffer and reservoir buffer (Zhang et al. 2013). When using a new ratio that contains more protein solution than precipitant solution, besides the protein molecular, more buffer reagents are also introduced into the drop. Then the protein buffer will predominant the final pH value, which can lead to an unrepeatable result for some pH-sensitive protein (Meged et al. 2008). On the other hand, if the new drop ratio uses more precipitant solution than that in the previous, the excess of precipitant accumulating in the drop will further exacerbate the over-nucleation (Dessau and Modis 2011; Vekilov and Vorontsova 2014). In contrast, adding additional deionized water to equally dilute the protein and precipitant concentrations has unique advantages that do not dramatically change the crystal drop final pH and create a lower supersaturated level to decrease the nucleate rate.

Spontaneous nucleation event can occur when the solution supersaturation degree crosses the supersolubility curve into the labile region (McPherson and Cudney 2014). The protein molecules overcome the energy barrier to aggregate as nucleus (Auer et al. 2007). Each nucleus can be a promising growth center to develop as a perfect single crystal with an orderly repeating pattern that is absolutely consistent with its crystal space group symmetry (Wukovitz and Yeates 1995). However, as the crystal grows, the well-arranged surface also can act as a potential nucleating point for additional growth (Vekilov and Alexander 2000). If a nucleus does form and has already developed as a crystal growth center, but the solution saturation level is still maintained in the labile region, at this time, the excess supersaturation pressure can become an archcriminal to force the solution protein molecule to secondarily nucleate on the existing crystal surface (Chernov 2003). For example, PDK1 PH domain in the current study is a vivid instance. In other cases, if the initial nucleation begins in a high-level supersaturation solution, the growth center is going to experience a longer time in the labile region, which also promotes the secondary nucleation event happen (Ahn et al. 2022). The crystal that comes from secondary nucleation usually has a completely different space group symmetry compared to its mother crystal. It is an important source of twinning (Thompson 2017).

Besides, it is worth mentioning that the protein concentration is not always completely the same within the drop (Sauter et al. 2015b). The region close to the growing center usually has a lower protein concentration than that in the remaining region of the drop due to the consumption of crystal growth that consistently recruits protein molecules from the nearby solution. In contrast, somewhere that far from the growing center can form a local high protein concentration region during crystal growth (Bergfors 2003). As the equilibration goes ahead, the protein concertation is undergoing an increase. Then the protein concentration fluctuation could happen in a non-linear way among different drop regions. If the crystal growth path is always upon a relatively high saturation level, occasionally, the protein concentration fluctuation would break through the supersolubility curve and reach the labile region, which often causes secondary nucleation at the new region (Saridakis and Chayen 2003).

Moreover, for the reason of gravity, the protein concentration around the drop bottom is prone to be slightly higher than that in the upper part after placing on the incubator for a long time (Moreno 2017). From another perspective, the plate surface also can act as a perfect crystalline surface for heterogeneous nucleation (Abyzov et al. 2019). Therefore, when the bottom high protein concentration meets with the solid plate surface it usually results in heterogeneous nucleation instead of homogeneous nucleation in the solution (Liu et al. 2023). Due to the direction limitation, the crystal nucleus that sticks on the flat plate surface can only develop upward or around, which often produces two-dimensional crystals with more lattice defects compared to the complete three-dimensional crystal. Many thin plate-shaped crystals produced from initial screening are typical examples. Adding additional water to dilute the high supersaturated level is helpful to offset the influence of gravity by increasing the density of the ingredient-derived flow that occurs between the drop edge and drop center (Nanev et al. 2004; Tanaka et al. 2013).

As for the above cases, adding additional deionized water into the drop to decrease the drop total supersaturation level is really necessary, especially for the protein whose metastable region is narrow but whose labile region is very large (Camino et al. 2021). The additional water can shorten the nucleation path from its starting point to the metastable region on the phase diagram (Fig. 4). In other words, an appropriately diluted crystallization environment allows nucleation events can start under a relatively low supersaturation level. It reduces the staying time in the labile region for the newborn crystal nucleus and prevents the supersaturated solution produce too many nuclei. Then a few nuclei that had traveled to the metastable region can gradually grow up as a well-organized crystal with a three-dimensional shape and sufficient size, simultaneously, containing less twinning and intermolecular disorder.

Figure 4. The schematic of protein redundant. In the two-dimensional crystallization phase diagram that only take the protein and precipitant concentration into consideration, when maintaining the protein concentration at the same level and increasing the precipitation concentration from A to B, the final equilibrium point will move from C to D along with the solubility curve. Although the initial levels are the same, the protein concentration can further decrease from C to E in the high precipitant condition. More protein molecules are salt out from the solution and used to form the crystals compared to that in the low precipitant condition. Therefore, the level of change between C and E is known as protein redundant in the crystallization trails that have fixed the protein concentration but different precipitant concentration. The distance from A to E is the total protein reserves during the crystal growth.

With the solution saturation level increase during equilibration, the protein maximum solubility limit decreases (Kramer et al. 2012). When the drop vapor diffusion pressure is close to the reservoir vapor diffusion pressure, the drop solution reaches its maximum supersaturation level, which also defines the maximum solubility limit of protein. The excess soluble protein molecules will dissolve out from the supersaturated solution through an organized pattern and form crystals to achieve a new balance. Until the protein concentration decreases to its maximum solubility limit or under that level, the protein molecules in the solution stop dissolving and the crystal ceases to grow (Russo Krauss et al. 2013). Based on the schematic phase diagram (Rupp 2015), we just define the amount of protein concentration decrement throughout the crystallization process as protein reserves, which can be calculated from the start point where the equilibrium is beginning to the endpoint in which the crystal stops growing. The crystal growth, from a nucleus to a microcrystal and then to a mature crystal, requires enough protein reserves as a support. Theoretically, a larger crystal requires more protein reserves. As fig. 4 shows, adding additional water can create protein redundant, which is equal to enhance the total protein reserves during the crystal growth.

Initial crystal screening by commercial kit usually just produces many microcrystals without sufficient size for x-ray diffraction (McPherson and Cudney 2014). When someone strives to grow a larger crystal and uses higher protein concentration as a starting point, sometimes, it is just the opposite of what one wishes. Especially for some sensitive and unstable proteins, even using the same precipitant concentration, a slightly enhanced protein concentration can bring the supersaturation level into the precipitation region. After the system precipitates the excessed protein molecules, the remaining protein starts nucleation and crystal growth again, just as before. But this time, only harvest the same microcrystal and additional precipitation, no larger crystal. To overcome this obstacle, adding additional deionized water at the same time could be a way out. Even though the protein concentration is increased to set up more protein reserves for follow-up crystallization, with the dilution of additional water, the initial protein concentration in the drop is still maintained in the labile region. As the crystallization goes on, there are more opportunities to grow larger crystals (Luft et al. 2011).

Crystal growth also needs to consume plenty of protein reserves. In particular with the existence of numerous grow centers, the soluble protein molecules in a supersaturated solution are expected to be consumed at a faster speed. However, the area of metastable region is limited for a given protein. The crystallization travel path on the phase diagram is also fixed in a setup precipitant concentration. When a large number of nuclei competitively recruit around protein molecules binding to its surface, the solution saturation state quickly pass through the metastable region to the stable region where crystals stop to growing further (Luft et al. 2011). As a result, the newborn microcrystal still does not have enough chance to become larger, the protein reserves have been depleted. The microcrystals are just microcrystals.

If the crystal grower adds an appropriate additional deionized water into the drop at the beginning point, the process of nucleation could be more controllable under a relatively low supersaturated level, which only produces a few nuclei as promising growing centers. In comparison to previous vicious competition within the masses of nuclei, these oligarchic growing centers can make better use of the limited protein reserves to grow up as a larger crystal.

According to current study, there is an optimal interval when adding the additional deionized water to help crystal optimization. In general, the effect of a small volume of additional water could be very limited, and no observed improvement. However, sometimes, the optimization could be failed by adding too much deionized water. The optimum volume requires the crystal grower to explore step by step. In the initial stage, it is highly recommended to gradually increase the additional water volume at a gradient as small as possible. If it shows effective feedback, raising the protein concentration also can be taken into consideration simultaneously. Anyway, optimizing these two factors by turns is likely to get a magical result.

As an extension, when adding the additional deionized water to optimize the co-crystal result, the exceeded ligand also can be dissolved into the deionized water and added into the drop together. As we know, the unfriendly property of ligands is one of the obstacles to successful co-crystallization, for example, some ligands whose solubility in water is really poor, or the ligands with low binding affinity (Hassell et al. 2007). In these cases, it is difficult to achieve a sufficient ligand occupancy ratio in the crystal, which usually leads to false positive results in co-crystallization (Müller 2017). In order to deal with above problems, the advantage of utilizing current method is obvious. It can significantly increase the ligand total amount in the drop but does not alter the final crystallization reagent ratio, which could be a more efficient way to harvest a real co-crystal. On the other hand, supplying ligands through the additional water can dramatically decrease the usage ratio of organic solvents, like DMSO, which chemical could affect normal crystallization in some ways.

Impurities adsorb on the crystalline surface and ultimately form an impurity adsorption layer that prevents further growth of the crystal (Caylor et al. 1999; Yoshizaki et al. 2004). The research of Plomp et al. (2003) proved that adding additional water to the crystal drop that already stops growing can temporarily remove the crystal surface deficiency, and then, restart the crystal growth. The underlying mechanism of that method also can be used to guide how to add additional water for microcrystal optimization. Besides adding additive water at the beginning point, it is also can be applied for the other stages during crystallization. One of the applications is to discontinue the nucleation. When adequate nuclei have already been produced, further nucleation activity should be suppressed. However, in most cases, the supersaturation level is still located in the labile region where nucleation is sustained. In order to shift the saturation state to produce fewer nuclei and grow crystals to a larger size, there is no better solution than adding additional deionized water into the solution. The other application is recrystallization when the crystals cease to grow. The impurity in protein solution often causes more or less crystal surface defects. When the surface defect accumulates to a certain degree, the crystal stops growing further (McPherson and Kuznetsov 2014). Here, adding additional deionized can draw back the solution supersaturation level, resulting in a certain extent dissociation of crystal surface protein molecule. For this reason, the impurity that already binding on the crystal surface also leaves away. The recovered crystal therefore can continue its growth.

Last but not least, current method also can be combined with the seeding technique. At present, seeding is a widespread application method, which can take advantage of the property of metastable region to grow high-quality crystals (D’Arcy et al. 2014). On the original seeding basis, adding additional water can be an improved method for crystal optimization. In general seeding practice, both the protein and precipitant concentrations need to be diluted to a level where spontaneous nucleation does not happen (Bergfors 2003; D’Arcy et al. 2003). On this account, the protein reserves are reduced distinctly. However, by adding additional deionized water, it is more convenient to create a metastable environment based on the original condition, which gives the seeds more chance to grow as larger crystals. Moreover, to avoid the dissolution of the seeds under a low supersaturated solution, cross-linking the seeds with glutaraldehyde in advance is also recommended (Luft and Detitta 1999).

Taken together, our result contextualized that adding additional deionized water has the potential to decrease nucleation rate and improve final crystal size, as well as exhibit better morphology than previous crystals, indicating that this method can be adopted as an optimization strategy for protein crystallization. The best ratio of additional water that is added to the drop, however, is dependent on protein to protein, which is needed to perform a series of trials. Hope our work can spur further study of this fascinating process and give a green light to more potential applications.

The authors declare that they have no conflict of interest.

This work was supported by the National Research Foundation of Korea (NRF), grants numbers NRF- 2021R1F1A1061607 and 2020R1A6A1A0304370812 and the Gachon University research fund of 2020 (GCU-08420006). We would like to thank the staff of Beamline 11C at the Pohang Accelerator Laboratory (PAL), Korea.