Protein crystal occurrence domains in selective protein crystallisation for bio-separation

Bio-separation is a key bottleneck in the manufacture of biopharmaceuticals. In this work, we report an experimental evidence of direct selective protein crystallisation from a binary protein mixture solution. Lysozyme-thaumatin mixtures with a wide protein composition range (0 – 100 mg/mL, respectively) were tested against the same crystallisation cocktail conditions using hanging-drop vapour-diffusion (HDVD) crystallisation method. This work demonstrates selectivity of crystallisation from a model binary protein mixture and four crystal occurrence domains were determined as the operation windows of selective crystallisation of the target protein: 1) unsaturated region with no crystal formation, 2 & 3) target region with only single type of protein crystals (lysozyme crystals only or thaumatin crystals only) and 3) mixture region in which a mixture of both types of protein. This study demonstrates that protein crystallisation is not only applicable to high-purity protein solution and this also emphases the vital impacts of the presence of protein impurities in the process of target protein crystallisation. The study concludes protein


Introduction
Therapeutic proteins such as monoclonal antibodies have attracted major interests in the current pharmaceutical market. Progresses in biotechnology in the past decades has resulted in the approval of 285 distinct active biopharmaceutical products which are predominated by monoclonal antibodies and the sales of these products continue to grow reaching a total revenue of $188 billion alone in 2017. 1,2 Nevertheless, downstream separation processes have now become the bottleneck of cost-effective manufacturing of these protein-based products. Up to 80 % of manufacturing costs attribute to the downstream purification processes which mainly rely on multi-step protein A chromatography technology. 3 In the meanwhile, advances in upstream processes such as cell technology have led to higher titres of secreted proteins which is now beyond 5 g/L, creating greater challenge for protein A chromatography as effective purification steps. 4,5 Researchers are seeking alternative technologies to replace or partially substitute conventional chromatography steps towards more lucrative, rapid, and robust downstream separation processes. Various alternative separation technologies have been investigated, 6 including solvent extraction, 7-10 membrane-based method, [11][12][13][14] precipitation [15][16][17] and crystallisation. 17,18 Crystallisation serves as a common purification 3 processes of both inorganic and organic small molecule products in numerous conventional chemical industries. Moreover, crystalline protein is believed to have a higher purity and stability which can benefit formulation, storage, and drug delivery steps. [19][20][21][22] Though most of the studies of protein crystallisation focused on obtaining large single crystals for biomolecular structural studies, in the past few decades, researchers have demonstrated that protein crystallisation is also a feasible and scalable purification and isolation technology for downstream bio-separation. Judge et al. demonstrated the feasibility of ovalbumin recovery via bulk crystallisation in 1 L stirred batch crystalliser. 23 Jacobsen et al. were able to obtain microbial lipase crystals from clarified concentrated fermentation broths. 24 Hekmat's group has demonstrated that crystallisation is a scalable process by successfully transfer crystallisation of antigen-binding fragment FabC225 from 10 µL vapour diffusion experiments to a 100 mL batch crystallisation process. 25 A few continuous crystallisation platforms have also been developed in lab-scale to demonstrate the potential of adapting crystallisation in continuous manufacturing fashion, including stirred classified product removal tank, 26 tubular plug-flow crystalliser, 27 continuous crystalliser with oscillatory flow, 28,29 and meso oscillatory flow crystalliser. 30 In most cases, the focus has been on the possibility of crystallisation from protein solution with high purity. A very limited cases of selective crystallisation were described.
Ghatak and Ghatak reported selective crystallisation from protein mixture realised by specific additives (salts) accompanied with charged nano-patterned surfaces. 31 A previous 4 study by our group reported that mesoporous nucleants with specific pore size, ordered structure, and narrow pore size distribution are able to promote selective protein crystallisation via controlling the nucleation process. 32,33 Still, in all the cases mentioned above the concentration range in this study was at a relatively low level which may not be applicable for fast purification purposes. Judge et al. studied protein crystallisation in the presence of protein impurities, 23,34 and, preferential separation of lysozyme from lysozyme -ovalbumin mixtures was achieved by seeded batch crystallisation. 35 Nonetheless, in the above cases, only a limited range of mixture composition was covered and the only one of the proteins in the solution can be crystallised out under the solution condition. Systematic knowledge of crystallisation behaviour of target protein from the mixture is still absent. Few information is available for introducing seeds/heterogenous nucleants and future scale-up to proceed to selectively crystallise target protein from a more complex solution. The reported preferential crystallisation cases only focused on specific scenarios either with ultralow protein concentration or with a very limited range of protein composition of the mixture.
In this study, we provide the first direct experimental evidence that bio-separation is practical via crystallisation from a binary protein mixture where both proteins are supersaturated and crystallisable under an identical crystallisation condition.
Potassium sodium tartrate tetrahydrate precipitant solution (28 mg/mL to 560 mg/mL) was prepared by dissolving potassium sodium tartrate tetrahydrate in 0.1 M PIPES buffer, pH 6.8. All precipitant solutions were filtered through 0.22 µm Millex-GS Syringe Filter Units (Millipore) before crystallisation trials. Protein (lysozyme/thaumatin) solution was prepared by dissolving the protein powder into the buffer solution which was the same buffer as used for precipitant preparation. Protein concentration in the solution was determined by Nanodrop One c microvolume UV-Vis spectrophotometer (Thermo Scientific™) at 280 nm using mass extinction coefficient (ε 1% ) of 26.4 L/g-cm for lysozyme and 12.7 L/gm-cm for thaumatin. 36 Protein solution with expected concentration higher than 100 mg/mL was diluted before measured by UV-Vis spectrophotometer. Lysozyme-6 thaumatin mixtures were prepared by mixing lysozyme solution and thaumatin solution with determined concentrations. All protein solutions were filtered through 0.22 µm syringe filter (VWR) before crystallisation trials.
HDVD crystallisation experiments were conducted in 24-well VDX™ plate with sealant (Hampton Research). Each well was filled with 500 µL precipitant solution as reservoir solution. A 4 µL droplet with equal volume of protein solution and precipitant solution (same as the reservoir solution) was deposited on a borosilicate cover glass (VWR). The cover glass with the protein-precipitant drop on was carefully inversed and sealed onto the well filled with reservoir solution. The crystallisation plates were then placed into the incubator (20 C ± 0.5 C). The plates were observed using CX41 optical microscope (Olympus) regularly after they were set-up. Microscopic images were captured using a GXCAM HICHROME-MET camera (GT Vision).
In this study, we determined the crystallisation results based on the observations of the droplets under the optical microscope. And the droplets were categorised into (1) no crystal, (2) precipitation, (3) only lysozyme crystal(s), (4) only thaumatin crystal(s) and (5) both lysozyme crystal (s) and thaumatin crystal(s). Due to the limitation of the maximum amplification of the optical microscope, only crystals larger than about 5 µm can be observed and the shape of the crystal can be recognised, i.e. result was marked as '(5) both types of crystals' providing at least one lysozyme crystal larger than 5 µm and at least one thaumatin crystal larger than 5 µm were observed in the droplet at the same 7 time. Considering the inherently poor reproducibility of protein crystallisation and the enormous crystallisation conditions tested, it was not feasible to exam every single crystal using characterisation techniques such as single crystal X-ray diffraction which can be low-throughput and excessively time-consuming to provide representative analysis for the whole sample population. Usage of microscopic images of the droplets is a fast and robust way to screen and track the crystallisation conditions in limited time and the results are real-time, relatively consistent and representative since all the droplets were examined rather than single crystals were sampled and tested off-line. To enhance the confidence level of our results, each condition was repeated at least 12 times to mitigate the inherently poor reproducibility of protein crystallisation due to the stochastic nucleation event. In the first run, experiments were repeated 12 times in 3 separate plates. If the same results were obtained from all the replicates, the condition was only repeated for 12 times.
In the second run, another 12 replicates of each condition were repeated. If 22 replicates were given the same results, providing about 95% confidence level, the condition would be only repeated 24 times. For the remaining conditions, more replicates would be tested up to 144 times. In general, conditions with protein concentration lower than 50 mg/mL were repeated in 48 to 144 droplets. Considering the higher degree of supersaturation of the proteins would reduce the fluctuation and uncertainties of the crystallisation results, conditions with higher protein concentration were repeated at least 12 to 24 times to reassure the accuracy of the results. 8

Determination of Protein Crystallisation Condition in HDVD Crystallisation
Experiments. The results in Table 1 show that lysozyme and thaumatin were able to be crystallised individually from their single-protein solutions against crystallisation condition in which potassium sodium tartrate tetrahydrate was used as precipitant. Yet, under the conditions investigated in this study, no thaumatin crystal was obtained by using sodium chloride as precipitant. The droplets remained clear or only precipitations were observed in the period of observation.  Figure 1C, both types of crystals are stained blue and thus 9 they were protein crystals. And the bipyramical shape thaumatin crystals possesses a deeper blue colour while the tetragonal lysozyme crystals possesses a lighter blue colour which may due to the different crystal mosaicity. Figure 1C reveals that both lysozyme crystals and thaumatin crystals can be crystallised out from a lysozyme-thaumatin mixture using the tartrate salt as the precipitant while still possessing distinct crystal shapes.
Thaumatin crystals remained as bipyramidal shape in the mixture. Lysozyme crystals were tetragonal crystals though defects might be detected under certain conditions. Therefore, further experiments where preferential crystallisation from lysozyme-thaumatin binary protein mixture was attempted were conducted by using potassium sodium tartrate as precipitant rather than sodium chloride.      when the initial lysozyme concentration was higher than 80 mg/mL, the number of thaumatin crystals increased. This increase may due to that lysozyme crystallisation was faster resulting from the high degree of supersaturation. Consequently, free lysozyme in the solution decreased and thus thaumatin crystallisation was less affected by lysozyme in the mixture. Another assumption is based on the nature of protein crystal that protein crystal retains relatively high solvent content comparing to small molecule crystals 37 .
Therefore, as more protein crystals formed from the mixture, less solvent was left in the mixture and thaumatin concentration might have increased accordingly.
Furthermore, as shown in Figure 2, when the initial thaumatin concentration was 10 - were chances that thaumatin did not crystallise when the initial thaumatin concentration was low.
In general, as shown in Figure 2, there was always one type of protein crystallised out from the solution first and then followed by the other protein crystals. And the sequence was decided by the composition of the mixture, i.e. degrees of supersaturation of the proteins. When enough time was provided, both lysozyme and thaumatin would crystallise out from the mixture. This suggests that when operation time was controlled properly, bio-separation can be achieved via preferential protein crystallisation even if protein impurity in the mixture was supersaturated and able to be crystallised out under the crystallisation condition.
We also suggest that the presence of another protein, acting as an impurity in the solution, will slow down the crystallisation process of both the target protein and the impurity protein itself. Still, the crystallisation process was not inhibited completely.
Additionally, in the model system studied in this work, we did not find evidence that the existing lysozyme protein crystal could stimulate thaumatin crystallisation or vice versa.
Hence protein crystals as seeding remains as an option to facilitate target protein crystallisation from the mixture without the risks of promoting the impurity crystallisation simultaneously.

Conclusion
In this study, we successfully demonstrated preferential protein crystallisation using lysozyme-thaumatin binary mixture as the model. There was no direct evidence in this study that protein solubility was changed due to the existence of protein impurity. In the model binary protein mixture, the presence of another