Packing bridges in protein crystal structures
aDepartment of Chemistry, University of Pavia, Viale Taramelli 12, I-27100 Pavia, Italy, bDepartment of Structural and Computational Biology, Max F. Perutz Laboratories, University of Vienna, Campus Vienna Biocentre 5, A-1030 Vienna, Austria, and cDepartment of Biochemistry, Faculty of Chemistry and Chemical Technology, University of Ljubljana, Aškerčeva 5, SI-1000 Ljubljana, Slovenia
*Correspondence e-mail: firstname.lastname@example.org
On the basis of a statistical analysis of the data deposited in the Protein Data Bank [Berman et al. (2000). Nucleic Acids Res. 28, 235–242; Bernstein et al. (1977). J. Mol. Biol. 112, 535–542], it is shown that two symmetry-related protein molecules are frequently bridged by a small molecule/monoatomic ion, which was used in the crystallization medium despite the fact that it is not a physiological ligand of the macromolecule. It is therefore sensible to suppose that some of the solutes used in crystallizations can favour the nucleation process by bridging and opportunely orienting adjacent protein molecules. This would explain why small changes in the composition of the crystallization solution, for example, the presence of a minor amount of a specific additive, can have a dramatic impact on the outcome of a crystallization experiment.
Contrary to most materials, proteins are reluctant to crystallize. In fact, it was suggested that proteins have evolved to avoid crystallization which might compromise the viability of the cell (Doye et al., 2004). Nevertheless, in vivo crystallization is actually known to be a rather common event and a few spontaneous crystallizations have been observed: for example overexpression of human IgGs has been reported in the endoplasmic reticulum lumen of Chinese hamster ovary cells (Hasegawa et al., 2011); viral protein crystals can form in HEK cells infected by adenovirus (Franqueville et al., 2008); crystalline inclusion bodies were observed in various circumstances, including the epididymis of the nine-banded armadillo (Edmonds & Nagy, 1973) and rabbit embryos (Daniel & Kennedy, 1978); and pathological crystal-containing inclusion bodies were observed in rabbit epithelial cells infected with Bacteroides fragilis (Westrin, 1996). Self-assembly of insulin in secretory granules, storage proteins in seed and enzymes within peroxisomes has also been documented (Doye et al., 2004). Recently, in vivo crystallization in combination with free-electron laser-based serial femtosecond crystallography has emerged as a promising vehicle in structural biology: nano-sized crystals of glycosylated cathepsin from Trypanosoma brucei were grown in vivo in transfected Sf9 insect cells, then they were isolated and their structure was determined with serial femtosecond crystallography (Koopmann et al., 2012; Redecke et al., 2013). However, protein crystallization remains one of the principal bottlenecks in protein crystallography and any strategy to overcome this obstacle is essentially and merely heuristic (McCoy et al., 1999). Hundreds of experimental conditions varying in pH, ionic strength, type and concentration of precipitating agents, as well as temperature and concentration of protein, must be systematically and/or randomly tested to find crystallization hits, and it is well known that (seemingly) minor modifications of the experimental setting can lead to completely different results, for example the appearance of crystals, the formation of insoluble amorphous precipitates or the formation of crystals in a different space group.
On the basis of a statistical analysis of the data deposited in the Protein Data Bank (PDB; Berman et al., 2000; Bernstein et al., 1977), we here show that two symmetry-related protein molecules are frequently bridged by a small molecule/monoatomic ion, which was present in the crystallization medium despite the fact that it is not a physiological ligand of the macromolecule.
An example is shown on CheY, the regulator of the chemotaxis response in Escherichia coli, where the phosphorylation of Asp57 is accompanied by a conformational rearrangement of the β4-α4 loop (residues 88–91) and of Tyr106. A sulfate ion (shown in green) is intercalated between two protein molecules (Fig. 1). One of them (shown in blue) is related to the other (shown in red) by the symmetry operation (1 − x, + y, − z) of the P212121 space group of the crystal. The sulfate ion is clearly non-physiological and far from any active amino acid. Its presence is due to the crystallization conditions [2.2 M (NH4)2SO4] (Simonovic & Volz, 2001). The sulfate ion makes a hydrogen bond with the side-chain N atom of Lys119 and with a water molecule, which is in turn hydrogen bonded to the side-chain N atom of Lys136. The latter is directly hydrogen bonded to the side-chain N atom of Asn59 of the neighbour and symmetry-equivalent protein molecule, which is also connected to the sulfate ion with a hydrogen bond involving the main-chain N atom of Asn62.
It is therefore reasonable to suppose that some of the solutes used in crystallizations can favour the nucleation process by bridging and opportunely orienting adjacent protein molecules. This can explain why small changes in the composition of solutes in crystallization solutions can sometimes cause dramatic differences. In other words, small molecules or monoatomic ions present in the crystallization medium might mediate the nascent crystal packing contacts by stacking at the interface between symmetry-related protein molecules in the crystal lattice. Like for the direct crystal packing interactions (Carugo & Argos, 1997; Janin & Rodier, 1995; Janin, 1997), it can be expected that little biological information can be provided by these particular contacts. Therefore, the present communication is limited to a descriptive analysis of the presence of atoms/molecules that bridge protein molecules in the solid state.
The coordinates of the protein molecules surrounding the reference molecule were computed and a heteroatom within 4.5 Å from the reference molecule and from its symmetry-related neighbours was considered to be a `packing bridge'. In fact, this atom bridges the reference molecule and its neighbour. Any type of heteroatom belonging to any type of heterogroup was considered, with the exception of water molecules. Symmetry-related molecules were generated around the reference molecule by applying all the symmetry elements and by translating from −2 to +2 along the three unit-cell axes.
All the crystallographic data were taken from the Protein Data Bank (Berman et al., 2000; Bernstein et al., 1977). Attention was limited to crystal structures at good resolution (not worse than 1.5 Å) and without missing atoms or residues. Only monomeric proteins containing only one chain per asymmetric unit were retained: in this way we ensured that all the packing contacts and packing bridges are not physiological but are, on the contrary, a mere solid state artefact. Identical or nearly identical proteins were removed by fixing a pairwise sequence identity threshold of 95%. A more severe cutoff (like for example 25%), which might have been used to eliminate more drastically the sequence redundancy, was unnecessary, since in the present study our attention is focused only on non-physiological features that are independent of protein sequences.
There are about 11 000 pairs of molecules that form crystal packing contacts in the crystal structures examined in this work. Nearly 1300 of these interactions occur with a packing bridge. About 11.5% of the solid state interactions between symmetry-equivalent molecules are therefore associated with the presence of a heteroatom that connects the two molecules that are adjacent in the crystal state. A similar conclusion is reached by computing the percentage of inter-protein crystal packing contacts that involve a heteroatom for each individual crystal structure and by averaging this percentage value over all the crystal structures: 11.1% (±0.5) of the interactions between adjacent and symmetry-related proteins are mediated by a heteroatom.
Although only one-tenth of the crystal packing contacts involve a packing bridge, nearly one-half (45%) of the protein crystal structures have at least one packing bridge (see Table 1). This is apparently independent of the crystal system, since similar percentages are observed in all crystal systems. The same is true when considering the ten most common space groups, with the exception of the space group P6522, which is, however, the less common of the ten (see Table 2).
Packing bridges are particularly frequent when the crystal packing contact is due to a screw axis. Although only 57% of the symmetry operations in the protein crystal structures examined here are screw axes, 68% of the packing bridges occur when the two protein molecules that are in contact are related by a screw axis. This might suggest the hypothesis that this particular type of symmetry relationship tends to be generated between two identical molecules by a small non-physiological ligand. No particular trends are observed for the other types of symmetry, which are involved in packing bridges as expected on the basis of their frequency.
Many different atom types can form packing bridges. Often the atom is carbon (45% of the atoms) or oxygen (39%). Much less common are nitrogen (4%), sulfur (3%) and chlorine (1%). Other atoms include several metal cations, ranging from zinc, sodium and calcium to magnesium (but not iron, potassium or copper). The small molecules that most frequently are found in packing bridges are sulfate (15%), glycerol (11%) and 1,2-ethanediol (7%), and the ten molecules most commonly observed in packing bridges are reported in Table 3. Hundreds of other types of small molecules are, however, seldom observed in packing bridges, including glucose, mannose, lactose, maltose, and acetic and formic acids. It is not possible to interpret these frequencies as propensities, since the distribution of atom types amongst the heterogroups present in the crystallization medium at the nucleation stage is unknown. In fact, the exact concentration of all the heteromolecules during the crystallogenesis should be taken into account in order to draw general conclusions about the chemistry of the packing bridges. The present analysis is therefore necessarily qualitative.
A visual survey of the packing bridges suggests that most of their interactions with the protein molecules are electrostatic, in the form of both ionic interactions and hydrogen bonds. This is obviously not surprising, given the electrostatic nature of the surface of globular proteins. Moreover, it is also necessary to consider the fact that the nascent non-physiological contacts occur during the nucleation process in an aqueous environment in competition with the interactions with water molecules. A more detailed, chemical classification of all the interactions involved in packing interactions would be not only extremely expensive, since it would require a systematic visual inspection of each structure in order to be reliable, but also inconclusive since the data set is likely to be not representative of the global crystallization process, as is explained above.
By rejecting all packing bridges where the small molecule that links the two proteins has an average B factor higher than the average B factor of the protein, our attention is focused on the packing bridges that are more rigid. Obviously, they are only a subset of all the packing bridges (about one-third). However, their main features are very similar to those described above. Packing bridges between two proteins related by screw axes are more frequent than expected; their chemical composition is roughly the same; and the small molecules/monoatomic ions that are frequently observed as the packing bridge are the same.
In this study, we show that crystal packing contacts between protein molecules in the solid state are quite frequently mediated by a small molecule or by a monoatomic ion that bridges the two protein molecules, especially when the latter are related by a screw axis.
Given that these interactions are not physiological, they in general receive little attention from structural biologists. Consequently, their occurrence is likely to be underestimated and it is probable that water molecules are in some cases erroneously placed instead of other heteroatoms (Mueller-Dieckmann et al., 2007). On the basis of the available data it is unfortunately not possible to draw a statistical framework that might be used to predict the chemical composition of the crystallization medium that should facilitate protein crystallization. The observed frequency of the packing bridges mediating crystal packing contacts suggests that small molecules/monoatomic ions that can bridge two protein molecules might be a driving force of the crystallogenesis. In fact, often the presence of minor amounts of additives in the crystallization solution can critically influence the crystallization propensity or success of a crystallization experiment. In this respect, it is interesting to observe that packing bridges are always associated with a direct interaction between the two protein molecules. By considering also that in most of the cases (95%) a packing bridge connects only two protein molecules and not more than two, we presume that random encounters of two protein molecules in solution may be stabilized by an exogenous small molecule that bridges them and behaves as a template for the nucleation of nascent protein crystals. Interestingly, this phenomenon resembles the role of the N- and C-terminal segments in the formation of nascent crystal packing contacts (Carugo, 2011): given their conformational flexibility, these polypeptide segments are prone to make contacts with adjacent molecules in solution and therefore favour the nucleation processes.
Our results based on statistical analysis of packing contacts in protein crystals show that small molecules or monoatomic ions often mediate these contacts, in particular when positioned on a crystallographic screw axis. Our data suggest that the presence of small molecules can be important for successful crystallogenesis, as they may act as critical bridges that mediate crystal contacts. The nature of information available for our analysis, for the moment, does not allow us to design `magic bullet' formulations for crystallization screens. It must, however, be remembered that previous efforts in this direction showed that polyvalent charged molecules may favour protein crystallization by creating intermolecular non-covalent crosslinks between adjacent molecules in the solid state (McPherson & Cudney, 2006). The results of the present study extend these observations and suggest the value of investigating novel crystallization strategies that take advantage of the formation of packing bridges.
This work was partially supported by the BIN-III project of the GEN-AU initiative of the Austrian Government. Research of KDj was supported by the Federal Ministry of Economy, Family and Youth through `Laura Bassi Centre for Optimized Structural Studies' project No. 253275. We thank Georg Mlynek (University Vienna) for fruitful discussions that initiated this work.
Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H., Shindyalov, I. N. & Bourne, P. E. (2000). Nucleic Acids Res. 28, 235–242. Web of Science CrossRef PubMed CAS
Bernstein, F. C., Koetzle, T. F., Williams, G. J., Meyer, E. F., Brice, M. D., Rodgers, J. R., Kennard, O., Shimanouchi, T. & Tasumi, M. (1977). J. Mol. Biol. 112, 535–542. CrossRef CAS PubMed Web of Science
Carugo, O. (2011). Protein Sci. 20, 2121–2124. Web of Science CrossRef CAS PubMed
Carugo, O. & Argos, P. (1997). Protein Sci. 6, 2261–2263. CrossRef CAS PubMed Web of Science
Carugo, O. & Djinović-Carugo, K. (2012). J. Struct. Biol. 180, 96–100. Web of Science CrossRef CAS PubMed
Daniel, J. C. & Kennedy, J. R. (1978). J. Embryol. Exp. Morphol. 44, 31–43. PubMed Web of Science
Doye, J. P. K., Louis, A. A. & Vendruscolo, M. (2004). Phys. Biol. 1, 9–13. Web of Science CrossRef
Edmonds, R. H. & Nagy, F. (1973). J. Ultrastruct. Res. 42, 82–86. CrossRef CAS PubMed Web of Science
Franqueville, L., Henning, P., Magnusson, M., Vigne, E., Schoehn, G., Blair-Zajdel, M. E., Habib, N., Lindholm, L., Blair, G. E., Hong, S. S. & Boulanger, P. (2008). PLoS One, 3, e2894. Web of Science CrossRef PubMed
Hasegawa, H., Wendling, J., He, F., Trilisky, E., Stevenson, R., Franey, H., Kinderman, F., Li, G., Piedmonte, D. M., Osslund, T., Shen, M. & Ketchem, R. R. (2011). J. Biol. Chem. 286, 19917–19931. Web of Science CrossRef CAS PubMed
Janin, J. (1997). Nat. Struct. Biol. 4, 973–974. CrossRef CAS PubMed Web of Science
Janin, J. & Rodier, F. (1995). Proteins, 23, 580–587. CrossRef CAS PubMed Web of Science
Koopmann, R. et al. (2012). Nat. Methods, 9, 259–262. Web of Science CrossRef CAS PubMed
McCoy, A. J., Fucini, P., Noegel, A. A. & Stewart, M. (1999). Nat. Struct. Biol. 6, 836–841. Web of Science PubMed CAS
McPherson, A. & Cudney, B. (2006). J. Struct. Biol. 156, 387–406. Web of Science CrossRef PubMed CAS
Mueller-Dieckmann, C., Panjikar, S., Schmidt, A., Mueller, S., Kuper, J., Geerlof, A., Wilmanns, M., Singh, R. K., Tucker, P. A. & Weiss, M. S. (2007). Acta Cryst. D63, 366–380. Web of Science CrossRef CAS IUCr Journals
Redecke, L. et al. (2013). Science, 339, 227–230. Web of Science CrossRef CAS PubMed
Simonovic, M. & Volz, K. (2001). J. Biol. Chem. 276, 28637–28640. Web of Science CrossRef PubMed CAS
Westrin, K. M. (1996). Ultrastruct. Pathol. 20, 563–569. CrossRef CAS PubMed
© International Union of Crystallography. Prior permission is not required to reproduce short quotations, tables and figures from this article, provided the original authors and source are cited. For more information, click here.