3.1. Acetylcholine is hydrolyzed in vapor diffusion crystallization set-ups
In vitro analysis of the binding characteristics of ChoX revealed a high affinity for choline (K
d 2.3 ± 1.0 μM) (Oswald
et al; submitted for publication). Surprisingly, not only the natural transport substrate choline is able to interact with the binding protein [
8]. Acetylcholine can also bind to ChoX with moderate affinity (K
d 100 ± 12 μM) (Oswald
et al; in preparation).
In order to reveal the binding properties of both ligands, ChoX was subjected to structural analysis. For this purpose ChoX was crystallized in a conventional vapor diffusion setup supplemented with an excess of the respective ligand. Independent of the supplemented ligand, protein crystals appeared within 4 weeks. However, in the case of acetylcholine structural analysis revealed only little electron density in the binding site. In fact the electron density obtained from the crystal grown in an acetylcholine supplemented setup more resembled the density obtained from the co-crystallization experiment of ChoX with choline. This finding suggested that the ligand situated in the binding site indeed is choline, obtained after the hydrolysis of acetylcholine. The crystals which were subjected to analysis were grown from a precipitant solution containing acetate buffer at a pH around 4.5. Furthermore, the crystals only appeared within one month. Therefore, it seems very likely that acetylcholine was hydrolyzed during the time the crystals needed to emerge. This assumption is supported by biochemical and structural work on acetylcholine esterase and the acetylcholine binding protein [
16]. The hydrolysis process is probably catalyzed by the acidic environment of the setup. The products of this hydrolysis, acetate and choline, further compete with acetylcholine for the binding site. As choline exhibits a higher affinity to ChoX than acetylcholine, the prevailing ChoX species in the setup will hold choline in the same binding site. This explains why the obtained crystals consisted of ChoX/choline complexes instead of ChoX/acetylcholine ones. Nevertheless, the acetate is diffused away in the crystallization drop and not found inside the crystal, suggesting that hydrolysis occurs before the initial crystallization nuclei is formed, although we can not completely rule out that hydrolysis can also occur within the crystals.
Figure 1A displays a final 2F
o–F
c electron density map (colored blue) contoured at 1 sigma, of the ligand in the ChoX/acetylcholine setup. For clarity, amino acid residues participating in ligand binding are shown. Here, the density was obtained from a crystal of a conventional vapor diffusion setup with acetylcholine at a resolution of 1.9 Å. However, the present density does not cover the entire acetylcholine ligand. Rather, the density strongly supported our notion that the acetylcholine was hydrolyzed during the time of crystal formation and only choline was bound to the protein. However, we cannot completely rule out the presence of a mixture of choline and acetylcholine in the crystals obtained by the conventional co-crystallization set-up. If this were indeed the case, these crystals would be unsuitable to determine the ChoX/acetylcholine structure.
3.2. Seeding for quick crystal growth
Co-crystallization of ChoX with acetylcholine in a conventional vapor diffusion setup was unsuccessful as the ligand was hydrolyzed during the time the crystal needed to grow. To circumvent this problem we anticipated that either crystallization of ChoX/acetylcholine at neutral pH values or accelerated crystal growth was required. Although different alternative crystallization conditions were tested, none of them produced suitable crystals of ChoX complexed with acetylcholine. In order to obtain crystals containing fully preserved acetylcholine acceleration of the crystallization process seemed crucial.
The microseeding method promised to be most successful for this purpose. This technique takes advantage of the fact that the formation of a crystal is a two step process divided into nuclei formation and crystal growth. The initial step, nuclei formation, is more likely to occur if the protein solution is highly supersaturated [
17]. In contrast, the growth of crystals, an ordered process, is maintained in the metastable zone of the phase diagram. Seeding methods separate the two events of nucleation and crystal growth [
18]. In microseeding this separation is accomplished by transferring a seed, a submicroscopic crystal, from one condition, where the level of supersaturation is high, to a similar condition at a lower level of supersaturation. In order to have lower levels of supersaturation either the protein or the precipitant concentration is lowered in a crystallization setup [
19].
As no crystals of ChoX complexed with acetylcholine were available a crystal of ChoX with its natural ligand choline was utilized to obtain seeds. Submicroscopic seeds, obtained by touching the crystal with a horse tail hair [
18], where transferred by streak-seeding [
20] to a freshly setup drop supplemented with acetylcholine instead of choline. As the crystallization of ChoX complexed with acetylcholine required rapid crystal growth, the seeding conditions were chosen in such a way that high saturation would be guaranteed. Therefore, the precipitant concentration was not lowered in comparison to conditions in which ChoX/choline crystals were obtained. Crystals of ChoX complexed with its new ligand grew along the streak seeding line, as exemplified in
Figure 2A. By utilizing the microseeding method it was possible to obtain initial, small crystals of ChoX with acetylcholine in less than 3 hours. These crystals grew to sizes suitable for data collection within 24 hours. After 24 hours these crystals were harvested and flash frozen in liquid nitrogen to prevent further hydrolysis of acetylcholine. A time course of the crystal growth process utilizing the microseeding method is depicted in
Figure 2B.
Astonishingly, the quality of these crystals was extremely high, showing diffraction up to 1.8 Å. The resulting electron density in the binding site is well defined for the acetylcholine molecule.
Figure 1B shows the electron density, colored blue, in the binding site after molecular replacement and one round of restrained refinement in Refmac5. Here, the same search model as in
Figure 1A was utilized for molecular replacement, meaning that no ligand was present in the binding site.
Figure 1A only shows little density, which is probably accounting for a partial occupation of the binding site with a choline molecule. In contrast, the electron density in
Figure 1B is much larger and of the defined shape of an acetylcholine molecule.
3.3. Is seeding inducing twinning?
To avoid hydrolysis during the crystallization process ChoX was crystallized complexed with acetylcholine utilizing a microseeding setup. Establishing microseeding resulted in very rapid crystal growth (less than 24 hours) of ChoX harboring the non-hydrolyzed substrate. Crystals of ChoX complexed with choline all exhibited space group P21. However, the unit cell parameters of the crystals obtained by microseeding experiments showed a β angle near 90°. Furthermore, the data set scaled equally well in an orthorhombic lattice, which was the first warning sign of twinning. In a monoclinic space group, twinning is rather uncommon, but might occur if the β angle is close to 90°, as it was the case in our study. Therefore, the data was carefully subjected to twinning analysis.
Figure 3 shows different statistical data analysis of reflection data derived from a crystal grown in a conventional vapor diffusion setup (without seeding) and a crystal obtained by microseeding (with seeding).
The Rees plots [
21] (
Figure 3A) of both datasets give a first indication of crystal twinning. Here, the fraction of local average intensity of the acentric reflections z is plotted against the cumulative distribution of z. The red curve shows the theoretic contribution in case of non-twinned data, which follows an exponential progression. However, the crystal obtained by seeding shows a lower amount of weak intensities. The more sigmoidal progression is another indication for twinning [
21]. Analysis according to Yeates not only indicates twinning in case of the crystal obtained by seeding but it also provides a first estimate for the twinning factor α [
22,
23].
Whereas the crystal grown by conventional vapor diffusion setup shows no twinning the crystal from the microseeding experiment shows a high twinning factor above 0.4 (
Figure 3B). A very robust test for twinning which also gives a good estimate for the twinning factor and is rather insensitive to anisotropic diffraction is an analysis using the L-function [
10] (blue curve in
Figure 3C). The analysis of the twinned crystal shows an extremely high value for α approaching almost 0.5, the value for a perfect twin (upper red curve). In contrast the crystal of the conventional vapor diffusion setup shows no twinning as its L-function follows the progression for untwinned data (lower red curve).
With the help of microseeding it was possible to acquire crystals within a few hours. However, these crystals showed a high twinning fraction in contrast to the crystals grown under conventional conditions. Therefore, the question arises wether there is a correlation between seeding and twinning. In a conventional setup, where nuclei need to be formed first, the nucleation process is guided by the nucleation barrier [
24]. In contrast, when a seeding experiment is conducted and nuclei are provided this nucleation barrier is of minor importance. Thus, a seeding experiment requires a lower saturation level than a conventional crystallization setup [
18,
20].
However, in order to drastically accelerate the growth process for ChoX with acetylcholine conditions supporting a very high level of saturation were chosen. Thereby, the equilibrium between the segregation of a solid phase (crystal) and the dissolving of the newly formed phase is also influenced. Thus, the new solid phase forms very rapidly. Unfortunately, this enhances the occurrence of crystal growth disorders, such as twinning, as possible defects in crystal development are less likely to be dissolved. However, this also means that crystals grown less rapidly possibly exhibit a lower degree of twinning.
This hypothesis was tested with a mutant crystal of ChoX. Here the crystal appeared 3 days after streak seeding was used to induce crystal formation and indeed showed a lower twinning fraction (α∼ 0.35). This supported our hypothesis that the twinning fraction in the case of the ChoX/acetylcholine crystals was indeed dependent on the time of crystal grows. However, one has to keep in mind that twinning was only possible due to a beta angle close to 90°. Thus, the strategy outlined here, demonstrates the power of microseeding to obtain X-ray suitable crystals of labile substrates. An analysis of twinning is easily performed nowadays and even in our case (perfect pseudo-merohedral twinning), the crystal structure could successfully be solved (Oswald et al. submitted for publication).