Several features of DNA or RNA play important roles in the process of co-crystallization. Apart from purity, the length and the ends (blunt or sticky) are also factors that must be explored when crystallizing nucleic acid-protein complexes.

Perhaps one of the most significant variables is the length of the nucleic acid. A general rule in protein-nucleic acid crystallization is to identify a sequence of minimal length that binds tightly to the protein [40,41]. On the one hand, oligonucleotides that are too short will destabilize the complex, as it might limit the number of potential interactions between nucleic acid and protein, but on the other hand it may improve crystal quality, as it will reduce flanking regions that could disturb crystal contacts. Consequently, determining the minimal binding sequence of an aptamer should be considered essential. Although this is not always straightforward and can be time consuming, removing nucleotides that are unnecessary for target binding will prevent them from potentially disturbing crystal contacts.

In addition to blunt-ended nucleic acids, single- or double-base overhangs (sticky ends) are commonly explored when crystallizing complexes containing DNA or RNA [40]. It is often observed in crystals that sticky ends form crystal contacts by base-paring with complementary sticky ends, forming a pseudo-continuous double helix. In order to allow the best end-to-end packing, the overhanging bases of one strand should be complementary to the overhanging bases of the opposite strand [41]. Nevertheless, this may not be easily applicable to aptamers, because sticky ends could disrupt the tertiary structure, and hence abolish binding, of the aptamer. It has been observed, however, that the 5′ extensions from two molecules of the vitamin B₁₂ RNA aptamer form a crystal contact by creating a six-base-pair duplex with two stacked adenosine–adenosine pairs [33] (Figure 1).

It may also be critical to have highly purified DNA or RNA to obtain well-ordered crystals of a protein-nucleic acid complex [41]. The most common methods for purifying synthesized oligonucleotides for use in crystallization experiments are anion exchange chromatography and purification from polyacrylamide gels (PAGE). Because results of purification can vary slightly, it is sometimes desirable to further treat the purified oligonucleotide, for instance by dialyzing against an appropriate buffer.

RNA synthesis is relatively expensive; if no modified nucleotides are required, it could be cheaper and perhaps more convenient to transcribe the RNA from a DNA template. When more stable RNAs are required, one could order chemically synthesized RNA aptamers, in which all or specific uridines and cytidines are 2′-fluoro-modified. In some cases, these modified nucleotides have been used during both aptamer selection and crystallization; e.g., [23,42].

A general requirement for crystallization experiments is that the protein needs to be homogeneous, i.e., highly pure (>98%) and in a low polydispersity state [43]. DLS is a commonly used technique to measure the polydispersity of a protein sample. Other parameters that should be established before setting up crystallization screens are the stability of the protein in different buffers, and whether the protein is correctly folded (CD spectroscopy) and active (activity measurements).

It is good practice to try several protein:aptamer molar ratios. In crystallization experiments, protein and DNA are usually mixed at 1:1.2 to 1:1.5 molar ratios with DNA in excess [41,43]. The principle behind this ratio is that the concentrations of protein and DNA are frequently rough estimations and DNA could possibly not saturate all binding sites on the protein. Our advice is, therefore, to use similar ratios when setting up crystallization experiments of protein-aptamer complexes.

Screens for obtaining crystals should include a wide variety of crystallization conditions. Nowadays, many commercial screens are available from companies such as Hampton Research, Emerald BioSystems, Molecular Dimensions, Jena Bioscience and Qiagen; these also include screens specific for crystallizing nucleic acids and nucleic acid complexes. The commercial screens are undoubtedly the easiest way to initiate the first crystallization trials. For more elaborate information on protein crystallization, see [38,43,44]. For more elaborate information on RNA and sample preparation for crystallization of RNA and RNA-protein complexes in particular, see [45–47].

In general, for nucleic acid crystallization, it is favorable to use polyethylene glycol (PEG) or 2-methyl-2,4-pentanediol (MPD) as precipitants, rather than high salt, because high salt may disrupt charged interactions between protein and nucleic acid. It is not possible to predict which screen will result in crystal formation, because many factors influence crystallization; variables include pH, ionic strength, temperature, protein concentration, the presence of various salts, ligands or additives, the type of precipitant and the crystallization method (hanging drop, sitting drop, dialysis, etc.) [38,44].

Thrombin is a trypsin-like serine protease with an important role in hemostasis, by converting soluble fibrinogen into insoluble fibrin strands. Shortly after the initial development of SELEX, thrombin-binding DNA aptamers were described [48], providing the first example of DNA binding to a protein normally not involved in DNA binding. The year after thrombin binding aptamers had first been described, a crystal structure of the thrombin-aptamer complex became available. It was obtained using a reservoir solution of 25%–30% (v/v) PEG 8000, 375 mM NaCl, 0.5 mM NaN₃ and 50 mM sodium phosphate at pH 7.3, which was also used for the crystallization of a thrombin-hirugen complex [21,49]. The crystal used for data collection grew in 2 months and diffracted to about 2.9 Å, it belongs to orthorhombic space group P2₁2₁2_1. In the same year, an NMR solution structure [50,51] became available for the thrombin-aptamer complex as well. The crystal structure and NMR solution structure revealed that the core of the thrombin-binding aptamers is formed by two stacked G-quadruplexes (Figure 2a), although the crystal and NMR solution structure show different topologies of the G-quadruplex [52]. It was also shown that the aptamer binds to exosite I of thrombin (Figure 2b). Recently, another crystal structure of a thrombin-aptamer complex has been published (crystallized at 20 °C using 20% (w/v) PEG 20,000, 200 mM ammonium sulfate, 3% (v/v) n-propanol, 100 M sodium acetate at pH 5.8). It contains a modified thrombin binding aptamer that binds thrombin with higher affinity [24].

Recently, the 2.15 Å resolution crystal structure of the anti-Fc RNA aptamer in complex with the Fc fragment of a human IgG1 antibody (hFc1) was elucidated [28]. In most cases, RNA aptamers bind to their target proteins predominantly via electrostatic interactions between the negatively-charged phosphate backbones and positively-charged surface residues of proteins. To this end, the interaction between this aptamer and hFc1 is an exception, as it mainly consists of van der Waals contacts and hydrogen bonds, rather than electrostatic forces. The structure also revealed other interesting features. The RNA structure in the complex diverges greatly from its predicted secondary structure; it changes its conformation to one that structurally fits to hFc1 (Figure 3). The aptamer–hFc1 interaction is stabilized by a calcium ion and, therefore, binding and release of the aptamer can be controlled by either chelation or addition of calcium.

Crystals of the RNA aptamer in complex with hFc1 were grown by sitting-drop vapor diffusion using a reservoir solution containing 0.1 M Tris–HCl buffer (pH 8.0), 20% (w/v) PEG 1000 and 0.2 M calcium acetate. [42]. The RNA aptamer was chemically synthesized containing 2′-fluoropyrimidines and purified by PAGE. NMR analysis showed that the interaction between the Fc fragment and the aptamer has a 1:2 (Fc fragment:aptamer) stoichiometry [53]. Consequently, the aptamer was mixed with the Fc fragment in a molar ratio of 1:2.2 (Fc fragment:aptamer) for crystallization. The crystals belong to the space group P2₁2₁2 and diffracted to 2.15 Å. Interestingly, crystal quality was improved by applying a solution stirring technique [54].

The sequence of an oligonucleotide, and the intramolecular base-pairing that this sequence enables, largely defines the 3D-structure of the oligonucleotide. Base-paired regions in the 3D-structure of the oligonucleotide are thought to act as stabilizers, allowing loops and bulges to position themselves in ways suitable for target interaction [56]. The nucleotide composition of distinct aptamers that bind the same target can vary strongly, particularly in the base-paired regions; however, the overall structure could remain largely similar [11,12,57]. In other words, structural elements shared between oligonucleotides might therefore not be easily visible from primary sequence alignments. In addition, defining the boundaries of the minimal sequence required for target binding is not straightforward. For these reasons, identifying a recurring structure, within a number of sequenced oligonucleotides of an enriched pool, can be challenging if only sequence alignments are used.

Secondary structure predictions of clones in enriched pools may reveal recurring structural features, at any position and with deviating primary structures of oligonucleotides. These predictions could also be useful tools for identifying the minimal binding sequence. Most popular secondary structure prediction programs, including mfold [58], may be of limited value as they can only predict relatively simple hairpin structures. For the analysis of more complex aptamer structures, dedicated programs should be used, e.g., for the prediction of pseudoknots [59] and G-quadruplexes [60].

Besides secondary structure prediction, software has been developed to predict tertiary structures, varying from ab initio modeling to approaches requiring detailed information on base pairing and other interactions [61]. Such software has not frequently been used in aptamer research, partly because these programs often do not provide a single prediction, but rather give numerous possibilities that each have to be scored for their relevance. For example, the MC-Fold and MC-Sym pipeline [62] that we have previously used for the generation of a 3D-model of a streptavidin binding aptamer, provided 57 widely different structural models [12]. Eventually a prediction was selected by comparing theoretical scatter of the predicted models with experimental SAXS data of the aptamer. A manual fit of the aptamer model onto a streptavidin crystal structure [12] showed that the aptamer could physically occupy two biotin-binding sites, and that it probably interacts with the loops that normally cover the biotin-binding pocket. This would explain the stoichiometry (2 aptamers per 1 streptavidin tetramer) and the presence of two distinct regions in the aptamer that are both essential for binding [12]. Although this modeling approach requires much more effort than standard secondary structure predictions, it can provide valuable information on the aptamer binding site and on the parts of the aptamer and the target protein that are involved in the interaction.

As mentioned above, one of the requirements for the initiation of crystallization experiments with aptamer-protein complexes is establishing the minimal binding sequence of the aptamer. As described, this can be relatively easy if a recurring sequence or structural motif is found, but binding should still be confirmed.

Alternatively, a full length aptamer may be shortened from either the 5′ or 3′ end to various lengths, by either enzymatic trimming [63] or alkaline hydrolysis [13,64]. The resulting pool of fragments of randomly distributed lengths is incubated with the target, bound and unbound fragments are recovered as separate fractions, and run on a gel (e.g., polyacrylamide) to separate all individual fragments. From the patterns that emerge on the gel and the full-length sequence, the minimal binding sequence can be deduced. An alternative approach to determine the minimal sequence required for binding is the use of fragments of defined lengths obtained by PCR [15] or as synthetic constructs [14], and screening them for their binding capacity, e.g., by SPR. To get a better understanding of the aptamer-target interaction, nucleotides that are predicted to be important, for instance based on sequence or secondary and tertiary structure modeling, could be mutated to other nucleotides and their effect on target binding determined.

Secondary structure predictions are particularly helpful to identify the smallest DNA or RNA molecule still capable of binding the target, and may also reveal the nucleotides that are involved in target binding. On the other hand, it is also useful to obtain information on which regions or amino acids of the target are actually interacting with the aptamer. To this end, alanine scanning, replacing specific amino acids in a protein with alanine, has been successfully used to show that two arginine residues in the protease domain of non-structural protein 3 from the Hepatitis C virus are essential for aptamer binding [65]. Although successful, this method is very laborious. Hydrogen/deuterium exchange in combination with mass spectrometry (HXMS) [66] could provide an alternative method to map sites on the protein that interact with the aptamer. Other examples of the investigation of ligand specificity are the binding of biotin to an RNA aptamer. Several small molecules, similar to biotin, or fragments thereof, were tested for their ability to bind the RNA aptamer. The results suggest that binding is not enabled by a single functionality, but rather that several interactions of different parts of biotin are required simultaneously [13]. Results of another study, in which the binding capacity of various compounds to an ethanolamine-binding aptamer was investigated, showed that the aptamer has a preference for only one functionality in the target molecule: either an ethyl- or methylamine group [67].

Other advanced analytical techniques are available for investigating aptamer-target interactions. ITC is a method used to determine thermodynamic parameters of interactions; the affinity constant can, however, also be deduced from ITC data. SPR and QCM can be used to measure the association and dissociation rates that underlie the affinity constant of an aptamer-target interaction. SAXS can be applied to obtain information on the size and shape of macromolecular structures; like SAXS, DLS (also referred to as Quasi-elastic light scattering) can also be used for the determination of particle sizes and shapes, but it also provides information on the diffusion coefficients in solution. CD is particularly useful in determining secondary structures of proteins and aptamer-protein complexes. Combining these dedicated analytical techniques can provide useful insights into the aptamer target interaction. A recent example for which a combination of these techniques was used (ITC, DLS, NMR and SAXS) is the transformation of a cocaine binding aptamer to one that became specific for deoxycholic acid, by only replacing a single base pair [19].