Optimizing the Growth of Endothiapepsin Crystals for Serial Crystallography Experiments

Optimizing crystal morphology
Step 1, optimizing crystal morphology, has been included to remind reader of its importance. It may be possible to create perfect micro-crystals from poorly diffraction needle-balls; however, the authors would suggest that it is better to optimize the two separately. First, find conditions that give rise to well-diffracting, single crystal via vapor diffusion, and then convert these conditions into batch rather than trying to do combine the two steps together. Discovering highly nucleating conditions, at this stage, is not necessary; morphology and diffraction quality are the principal goals.

Before beginning the micro-crystallization of endothiapepsin, an analysis of deposited structure crystallization conditions from the PDB was conducted. Crystallization conditions and approximate protocols could be obtained for 47 of the 48 depositions of enthothiapepsin. These were broadly all based upon the first crystallization of endothiapepsin conducted by Moews and Bunn (1970)⁴⁶. Given the similarities of these conditions and their 'classical' origin, a 96-well, vapor diffusion, sparse-matrix screen was performed to explore a wider variety of crystallization conditions. Endothiapepsin was concentrated to 70 mg/mL and a PACT sparse-matrix screen⁴⁷ was performed in a 96-well sitting-drop plate at 20 °C mixing 100 nL of protein with 100 nL of well solution. Every condition from this experiment after 36 h gave rise to crystals. However, an analysis of the crystal morphology indicated that some conditions might prove better for micro-crystallization optimization.

Figure 4A shows a drop from the PACT screen that was broadly representative of those observed in the majority of the plate. At first glance, it may be tempting to think that these crystals might be worth optimizing further for micro-crystallization. The crystals are large and there appears to be significant nucleation. However, the overall crystal morphology is not ideal. Firstly, the crystals are not observably singletons as it appears that multiple crystals are growing from single nucleation points. Secondly, the crystal size is highly asymmetric with growth principally occurring down a single axis. Such crystals are theoretically more likely to preferentially align when delivered to the X-ray beam. Both characteristics present problems during collection and processing of serial crystallographic data.

Figure 4B, however, shows endothiapepsin crystals grown in the presence of MgCl₂. This morphology was consistent across all conditions that contained MgCl₂ and therefore suggested that their morphology was due to MgCl₂. The MgCl₂ conditions produced single, more box-like crystals that represented a better target for the ultimate serial experiments.

There were four conditions within the PACT screen that contained MgCl₂. To better understand the influence of all the different components of these conditions on endothiapepsin crystallization, a random optimization was performed. A screen was created containing a random combination of the buffers and precipitants at a range of concentrations and pHs. The MgCl₂ concentration was also varied and then the resulting drops were arbitrarily graded from 0-5 (0 being no crystals or precipitation) in terms of their visual crystal quality and precipitation level.

Figure 5A shows a heatmap of the results from a Pearson's correlation analysis between the precipitation level and crystal quality, and the screen variables (examples of the drops from this experiment are shown in Figure 5B, C and D). The results indicated that the pH of the solution was highly correlated to the level of precipitation, with alkaline buffers resulting in more precipitation. MgCl₂ concentration was slightly correlated to the level of precipitation, as was the pH and precipitant concentration to crystal quality.

Based on these results, the decision was taken to take the crystals grown in 0.1 M Tris-HCl pH 7.0, 0.15 M MgCl₂, 20% (w/v) PEG 6,000 to the next step of the protocol - Transitioning to batch. The morphology of crystals was acceptable and an analysis of the X-ray diffraction and data quality metrics from these crystals suggested that there was no significant difference between the crystals grown in and out of the presence of Mg²⁺ (Figure 9).

Transitioning to batch
For many serial crystallography micro-crystallization optimizations, Step 2 will be the starting point. The protein of interest will have already been crystallized for cryo-crystallography and the crystallization protocol will now need transforming to create micro-crystal slurries. This protocol has only used 96-well vapor diffusion plates to perform the transformation to batch since vapor diffusion is the crystallization method used by 95% of PDB entries²⁶. The protocol has avoided moving into microbatch^34,35,37 since this transition might still incur a similar optimization. This is not to say that this protocol can only be done in vapor diffusion plates. All of the steps presented, would also work in microbatch if this was the original crystallization method.

To assess the crystallization of endothiapepsin in the chosen condition, a morphogram – or a rough phase diagram – was created. The purpose of the morphogram experiment is threefold. Firstly, an analysis of the morphogram is of great utility when assessing scaling routes in Step 3 - Scaling. Secondly, the morphogram acts as an optimization tool, helping to discover vapor diffusion conditions that give rise to crystals via batch [i.e., rapidly appearing crystals (< 24 h)]. Thirdly, if crystals have not appeared rapidly, an analysis of the seeded drops can give the crystallographer an idea of the approximate location of the current condition on the phase diagram. For example, if the seeded conditions give crystals but the unseeded do not, those conditions are likely to be in the metastable region.

The morphogram experiment of endothiapepsin was performed based on the 0.1 M Tris-HCl pH 7.0, 0.15 M MgCl₂, 20% (w/v) PEG 6,000 condition. The protein and PEG concentrations were varied from 100 to 12.5 mg/mL and 5 to 40% (w/v), respectively. The drops were analyzed and results plotted using the worksheet provided (Figure 6A).

It was also already clear from the Optimizing crystal morphology stage that endothiapepsin crystal growth in this condition, and at these protein concentrations, would result in crystals grown in under 24 h. This indicated that crystallization was occurring via a batch rather than a vapor diffusion driven process. The crystal grown in these conditions were, therefore, suitable for scaling to larger volumes.

If crystals had not been visible in the unseeded-drops after 24 h, then it would have been likely that crystallization was still dependent upon a transition (Figure 1B) and, therefore, not batch. In this case, the results from the morphogram experiment are still of interest. They give an indication of the probable starting point for crystallization on the phase diagram and hence, how the subsequent optimization should proceed. Look at the seeded drops. The seeds will allow for crystal growth in the metastable region regardless of nucleation. For example, if crystals appear within 24 h in the seeded-drops but not the unseeded drops, this indicates part of the metastable region can be observed. If no crystals are observed in either the seeded or unseeded-drops, all wells remain undersaturated.

Scaling
Looking at the morphogram (Figure 6A), a number of observations could be made. The amount of nucleation did appear to be affected by both the protein and precipitant concentrations. There was also a very clear demarcation of drops that lead to protein precipitation, with drops either containing: nothing, crystals or precipitate (Figure 6B). The addition of seeds (Figure 6D) also greatly increased X_n when compared to the drops without seeds (Figure 6C).Taking all these results together, it was decided to attempt to scale both a batch and seeded-batch protocol at 30% (w/v) PEG 6,000 and 100 mg/mL endothiapepsin.

The initial test scaling was done in 24-well hanging drop plates. The drop volumes were gradually increased so that any changes in crystallization behavior could be observed (Figure 7). As can be seen, in both the unseeded and seeded drops crystal growth has occurred. All the unseeded drops grew a range of crystal sizes, but predominantly large crystals (100-200 µm – longest dimension). The seeded drops, however, produced smaller crystals (5 – 50 µm – longest dimension). These initial tests suggested that seeds would be required to decrease X_s, but also, that this condition should be suitable for larger volumes.

When the volume was increased in 200 µL, the crystallization volume was continually agitated during crystal growth. The principal reason for this agitation was to ensure that the crystallization solution remains homogenous and that growing crystals do not settle on the bottom or sides of the tubes. Settling of crystals can lead to a heterogenous crystal population with both very large and small crystals. Agitating the crystallization solution can also promote nucleation^44,45.

Unfortunately, the unseeded 30% (w/v) PEG 6,000 produced no crystals, so the PEG concentration was increased to 35% (w/v). This increase improved the crystallization markedly, with a final X_n and X_s range of 3.6 ± 1.2 x 10⁶ crystals·mL^-1 and 42 ± 4.1 µm, respectively (Figure 8A and B – black). Although a significant improvement and an acceptable crystal concentration, the final crystals were too large for the planned experiment, so further optimizations were undertaken. To reduce the size of the final crystals two avenues were explored (Figure 1E): decreasing the protein concentration to try and limit the final crystal growth (Figure 8A and B – hot pink), and increasing the PEG concentration to try and increase nucleation (Figure 8A and B – green).

The reduction of the protein concentration unfortunately also dramatically reduced the X_n, which ultimately produced even larger crystals. Increasing the PEG concentration to 40% yielded a final X_n and X_s range of 3.1 ± 0.7 x 10⁶ crystals·mL^-1 and 39 ± 2.3 µm, respectively. These were not significantly different to the 35%, but since the final crystal size was reduced, this condition was continued with the further optimizations.

To increase the X_n, seeds were added. This dramatically increased the X_n (1.1 ± 1.8 x 10⁸ crystals·mL^-1) and lead to a smaller X_s (4.2 ± 4.0 µm) (Figure 8A and B – dashed purple). These crystals, although very suitable for some serial crystallography experiments, were deemed too small so the concentration of the added seeds was changed.

This tuning of the added seed stock, however, proved difficult to reliably repeat; therefore, quenching was attempted. After the addition of a seed stock, the crystal size was monitored and once a suitable crystal size was achieved (approximately 10 – 20 µm), the batch crystallization was quenched (Figure 8C and D). Quenching was proposed, with regard to micro-crystallization, in Kupitz et al. (2014)²⁵. Although perhaps not an ideal method, as protein solution will ultimately be wasted²⁶, the technique was very useful in this situation as crystal growth was difficult to control. The idea behind quenching is to rapidly return the crystallization mixture to a point just above the solubility line (Figure 1F). Once the solution has returned to the solubility line, the solution has returned to a stable saturated solution and no further crystal growth will occur.

Attempting to quench a crystallization reaction is not without risk. If too great a quenching solution is added, the protein in solution might be diluted so much that the solubility line is passed. In this case, the solution will become undersaturated and the crystals will start dissolving. To prevent this, it is possible to estimate the amount of required quenching solution based on the morphogram results. At the point of quenching, take the concentration of the protein solution. By comparing the protein concentration at the solubility line and the protein concentration in solution, an estimate of the required dilution can be made.

The quenched version of the 40% (w/v) PEG 6,000, 10 x diluted seed experiment gave a final crystal concentration and size range of 2.6 ± 3.1 x 10⁶ crystals·mL^-1 and 15 ± 3.9 µm, respectively.

Throughout the entire process, test X-ray data collections of the endothiapepsin crystals were collected at the Swiss Light Source PXII beamline using a 10 x 30 µm focus, an energy of 12.4 keV attenuated by 80%, and under cryo-conditions. The data were processed using dials and Figure 9 shows a comparison of CC_½. No dramatic change in in CC_½ was observed over the course of the optimization.

Figure 1: An overview of transitionary and batch crystallization, and scaling methods mapped onto a phase diagram. A. The zones and limits of the archetypical protein crystallization phase diagram. The precipitant and protein concentrations are plotted on the x and y axes, respectively, with the pure water point at the origin. The purple line indicates the protein supersaturation boundary, and the metastable, nucleation, and precipitation zones are shown in blue, green and pink, respectively. B. An example of the nucleation zone penetration limits of a 'transitionary phase' crystallization method, such as vapor diffusion. In this theoretical experiment, the drop precipitant and protein concentrations begin just below the solubility line – not yet supersaturated. While the drop equilibrates, the drop component concentrations increase such that the drop becomes supersaturated, and continues to move – or transition – into the nucleation zone. Upon crystal nucleation, the protein concentration in solution begins to drop. The concentration continues to fall as crystals grow until finally stopping at the solubility line. The blue dotted line marks a theoretical limit of the transition into the nucleation zone. As soon as nucleation begins, the protein concentration will drop, preventing further penetration. C. Example batch and seeded-batch crystallization trajectories. In batch, the mixing of the protein and precipitant must create a supersaturated solution within the nucleation zone so that crystal growth can occur. In seeded-batch, it is not strictly necessary to be in the nucleation zone due to the addition of micro-seeds, so locations in the metastable region can also be explored. D. A hypothetical optimization of the crystallization experiment shown in B from vapor diffusion to batch. The original vapor diffusion starting point has transitioned, via the resultant optimization vector, to the new start position; inside the nucleation zone. The resultant vector is the product of two optimizations: an increase in both protein and precipitant concentrations. E. Example optimizations when scaling batch conditions to tailor the final X_n and X_s. F. Quenching the crystallization experiment by the addition of crystallization buffer. It is essential that the quenching does not take the protein concentration out of the metastable region and, therefore, below the point of protein supersaturation. Otherwise, crystals will start to dissolve back into solution. B. and C. have been adapted from Beale et al. (2019)²⁶ with the permission of the authors. Please click here to view a larger version of this figure.

Figure 2: Increasing X_n and decreasing X_s. The idealized relationship between the number of crystals produced from a crystallization experiment and their mean longest dimension. To create this graph, the crystallization of a hypothetical 10 kDa model protein was used. The protein crystallized at a concentration of 10 mg/mL and yielded P2₁2₁2₁ crystals with dimensions of 49x50x51 Å. Every nucleation event was assumed to yield a crystal. Crystal growth was assumed to be homogeneous from every face. Please click here to view a larger version of this figure.

Figure 3: A flow-chart showing the steps to optimize a crystal grown in a small-volume (<500 nL), vapor diffusion experiment into a large-volume (> 100 µL) batch experiment. Crystal optimization is divided into three stages: (1) Optimizing crystal morphology. (2) Transitioning to batch. (3) Scaling. In Stage 1 it is important to identify suitable crystals for micro-crystallization. Some proteins only present in a single crystal morphology regardless of the crystallization condition. However, it is worth looking for conditions that give rise to single, cube-like crystals, or as close to these as humanly possible. Single, cube-like crystals, hypothetically and anecdotally, will generally give rise to better outcomes from serial crystallography experiments. Once a crystal morphology has been selected and the diffraction confirmed, it is then necessary to move the crystallization experiment from vapor diffusion to batch (Stage 2). Here, crystals should be optimized by their nucleation time. The goal is to find conditions that yield rapidly appearing crystals (> 24 h) as these conditions are likely to hit the nucleation zone immediately, and are therefore batch. Once a condition in the nucleation zone has been found, a morphogram can be created. The morphogram allows for the majority of the nucleation zone to mapped and potential scaling routes identified for Stage 3. The volume of an identified batch condition can then be either gradually or rapidly scaled in size to yield a final volume of >100 µL. Please click here to view a larger version of this figure.

Figure 4: An analysis of endothiapepsin crystallization conditions from a PACT sparse-matrix screen. A. and B. are photos after 24 h of wells A4 and C10, respectively from the PACT screen. The crystallization buffer components are highlighted on the figure. The SPG buffer is succinic acid, sodium dihydrogen phosphate, and glycine mixed in a 2:7:7 molar ratio. Please click here to view a larger version of this figure.

Figure 5: An analysis of the endothiapepsin crystallization optimization from the PACT MgCl₂ conditions. A. A heat map of the results from a Pearson's correlation analysis between buffer pH, MgCl₂ concentration, and precipitant concentration and the precipitation level and crystal quality. The precipitation level and crystal quality were both assessed arbitrarily on a scale of 0-5 (with 0 being no crystals or precipitation) after 24 h. B. C. and D. show examples of the crystallization and precipitation in three different drops. The crystallization condition and assessments of the precipitation level and crystal quality are also shown. Please click here to view a larger version of this figure.

Figure 6: An endothiapepsin morphogram when crystallized in 0.1 M Tris-HCl pH 7.0, 0.15 M MgCl₂ and PEG 6,000. A. A morphogram created from the "phase-diagram-generator" spreadsheet provided. The relative number of crystals in each drop is denoted by the size of the circles, and the results from drop 1 (protein and precipitant) and drop 2 (protein, precipitant and seeds) are highlighted in green and blue, respectively. The values of the protein and precipitant concentrations, on the x and y axis, respectively, denote pre-mixed values of each rather than final volumes. Based on the results, black lines and a purple line have been drawn to show the boundaries of the nucleation zone and metastable zone, respectively. B. C. and D. show some example results from the experiment. The red and blue dots marked on A. indicate the locations of B., and C. and D., respectively. Please click here to view a larger version of this figure.

Figure 7: Initial scaling trials of endothiapepsin in 24-well hanging drop plates. The same protein and precipitant concentrations were used for all trails: 100 mg/mL endothiapepsin in 0.1 M Na Acetate pH 4.6 and 0.1 M Tris-HCl pH 7.0, 0.15 M MgCl₂, and 30% (w/v) PEG 6,000, respectively. All of the displayed images were taken after 24 h and the final drop volumes are labelled on each image. The left panel (A, D, and G) are a 1:1 mix of protein and precipitant, the middle panel (B, E, and H) are a 1:2:3 mix of seeds, precipitant, and protein and the right panel (C, F, and I) are magnified images of the middle panel. Please click here to view a larger version of this figure.

Figure 8: Analysis of the endothiapepsin micro-crystallization in 200-300 µL volumes. A. and C. show how X_n changed over the experiment time. B. and D. show how X_s (longest dimension) changed over time. The results of the experiments have been separated for clarity. The red dotted line on C. and D. show the point at which quenching was performed. Please click here to view a larger version of this figure.

Figure 9: CC_½ results and images of crystals obtained at each stage of the micro-crystallization process to assess diffraction quality. A. CC_½ plotted against resolution from data collected from crystals grown: with and without Mg – part of the Stage 1 optimization, in a 200 nL volume, a 10 µL volume and the final 300 µL volume. B.C. and D. show the crystals from the 200 nL, 10 µL and 300 µL volume, respectively. Please click here to view a larger version of this figure.

Table 1: An example of the sample requirements for a hypothetical optical pump-probe experiment performed using fixed-targets. The protein used in this theoretical experiment was endothiapepsin. The fixed-target parameters were based on experiments reported in Ebrahim et al. (2019)⁴⁸ and Davy et al. (2019)⁴⁹. The sample information came from the protocol reported in this video article and the experimental variables were conservative estimates based on lived experience. The following sample requirements were subsequently calculated given the previous assumptions.