research papers
Human insulin
upon ligand binding and pH variation: the case of 4-ethylresorcinolaSection of Genetics, Cell Biology and Development, Department of Biology, University of Patras, GR-26500 Patras, Greece, bDiabetes Protein Engineering, Novo Nordisk A/S, Novo Nordisk Park, DK-2760 Malov, Denmark, cPANalytical B.V., Lelyweg 1, 7602 EA Almelo, The Netherlands, dEuropean Synchrotron Radiation Facility, CS40220, F-38043 Grenoble CEDEX 9, France, and eExcelsus Structural Solutions, Belgium
*Correspondence e-mail: imargiola@upatras.gr
This study focuses on the effects of the organic ligand 4-ethylresorcinol on the
of human insulin using powder X-ray crystallography. For this purpose, systematic crystallization experiments have been conducted in the presence of the organic ligand and zinc ions within the pH range 4.50–8.20, while observing crystallization behaviour around the of insulin. High-throughput crystal screening was performed using a laboratory X-ray diffraction system. The most representative samples were selected for synchrotron X-ray diffraction measurements, which took place at the European Synchrotron Radiation Facility (ESRF) and the Swiss Light Source (SLS). Four different crystalline polymorphs have been identified. Among these, two new phases with monoclinic symmetry have been found, which are targets for the future development of microcrystalline insulin drugs.Keywords: powder diffraction; human insulin; diabetes; synchrotron radiation; X-ray diffraction; pH variation; 4-ethylresorcinol.
1. Introduction
Diabetes is rapidly reaching epidemic proportions, affecting 150 million people worldwide and being projected to double in prevalence by 2025 (Zimmet et al., 2001; Carulli et al., 2005; Mogensen & Zimmet, 2002; Emami-Riedmaier et al., 2015). Since many cases go undiagnosed, these figures are likely to be an underestimate of its true prevalence. Left uncontrolled, diabetes can lead to coronary heart disease, kidney failure, blindness, limb amputations and premature death. The hallmark characteristic of type I diabetes is a lack of insulin.
Insulin consists of two distinct chains (A and B) which are linked together by two disulfide bonds. There is an additional intra-chain linkage in the A chain (Ryle et al., 1955). All three disulfide bonds are essential for the receptor-binding activity of insulin (Chang et al., 2003). Insulin in its monomeric form is an active hormone. However, the molecules tend to form dimers and in the presence of Zn2+ ions they form hexamers. The hexameric form is not active and acts as a storage form which provides the organism with the hormone when required. The addition of allosteric ligands (for example phenol and chloride ions) to insulin compositions is widely used to modify the pharmacodynamics and stability of pharmaceutical preparations (DeFelippis et al., 2001). Different formulations of the hormone are absorbed at different rates and have varying durations of action.
Crystallization has always been a key activity since the protein is often administered by the subcutaneous injection of crystalline formulations. Microcrystalline insulin solutions are already widely used in pharmaceutical formulations because of their stability and prolonged action. These pharmaceutical compositions usually contain insulin, zinc and a phenolic binder. Phenol results in further stability of the et al., 1992). Microcrystalline drugs exhibit certain advantages over formulations in solution. Higher concentrations of the drug can be achieved in crystals compared with the amorphous form in solution. Additional advantages include low viscosity of the composition and controlled release of the protein as the crystals gradually dissolve in the body (Basu et al., 2004). Crystallization of proteins is also less costly than lyophilization (Collings et al., 2010). Of course, the crystal supernatant should only contain additives that are approved as nontoxic. When packaged in a crystalline form, proteins display greater stability and resistance to chemical changes and are less sensitive to of their three-dimensional structure. Finally, crystalline proteins are often protected against proteolytic enzymes (Halban et al., 1987).
of the protein (BrangeAn important aspect that affects crystallinity is accurate control of the pH, as previous studies have reported that crystal morphology is often pH-dependent (McPherson, 1985; Farr et al., 1998). Protein solubility reaches a minimum at the (pI) and increases at both lower and higher pH values. The probability of yielding high-quality protein crystals, in terms of morphology, has been suggested to increase around the pI owing to the minimum protein solubility. However, numerous studies have shown that this is not always the case (Farr et al., 1998; Kantardjieff & Rupp, 2004). On the other hand, pH ranges, which are different for each protein, do exist where either lysine and arginine side chains begin to lose their positive charge or where, in an alternative case, the carboxyl groups of aspartic and glutamic acid side chains begin to lose their negative charge (McPherson, 1995). This partial neutralization of the molecule disrupts the formation of salt bridges between protein molecules and thus decreases the crystallization rate. A lower degree of nucleation is likely to result in fewer but larger and better-formed crystals owing to the control of rapid crystal growth at low and high pH.
Over the years, insulin has been crystallized and characterized in a number of crystal systems. Thanks to the three-dimensional insight obtained from dozens of crystal structures of the wild type (Hodgkin, 1971), mutants (Whittingham et al., 1998) and complexes with zinc ions and small molecules such as phenol-based ligands (Derewenda et al., 1989; Von Dreele et al., 2000; Norrman, 2007; Karavassili et al., 2012; Margiolaki et al., 2013; Valmas et al., 2015), it has been possible to fine-tune the kinetics of insulin dissociation. The resulting availability of a variety of insulin preparations with rapid or prolonged action profiles has improved the quality of life of millions of people (Brange, 1997). In the presence of Zn2+ ions, insulin crystallizes as a hexamer consisting of three dimers related by a crystallographic threefold axis (Bhatnagar et al., 2006; Norrman, 2007). The pharmaceutical formulations used for treatment are typically mixtures of crystalline and amorphous protein, which are injected subcutaneously, resulting in long-term hormone action owing to the slow dissolution of the protein crystals (Norrman & Schluckebier, 2007). Thus, insulin has been crystallized under several conditions in order to determine how the morphology of the crystals affects its release into the bloodstream.
To date, several different polymorphs have been identified and most of them are found to belong to the monoclinic, rhombohedral, tetragonal and cubic symmetries. Besides this type of 6, T3R3f and R6 depending on the type of B chains that they contain (Bhatnagar et al., 2006; Frankaer et al., 2012). The B-chain conformation depends on the zinc and chloride ion content of the crystallization solution, as well as that of other ligands (Adams et al., 1969; Bentley et al., 1976; Derewenda et al., 1989). Typically, in the absence of high chloride ion concentrations or phenolic derivatives the T6 hexamer is produced, while at high chloride or thiocyanate concentrations the T3R3f hexamer is produced (Smith et al., 2000). However, the addition of phenolic derivatives such as phenol or resorcinol will drive the transformation from the T to the R state, resulting in monoclinic crystals containing R6 hexamers (Derewenda et al., 1989).
different conformations of the B chain have been found, which subsequently lead to different conformations of the hexamer. Hence, three different hexameric conformations occur, which are denoted TIn this study, we investigate the effect of the resorcinol-based ligand 4-ethylresorcinol on the crystallization of human insulin (HI) as a function of pH. Resorcinol and its derivatives are extensively used as antiseptics and disinfectants in pharmaceutical formulations. The crystal ; Margiolaki, 2016; Karavassili & Margiolaki, 2015). Thereby, characterization of the insulin polymorphs produced based on this specific resorcinol ligand was performed.
must be fully characterized in order to produce a drug in a crystalline form. In this case, X-ray powder diffraction (XRPD) is the most appropriate tool for the characterization of the various polymorphs, providing information on the microcrystalline samples (Margiolaki & Wright, 20082. Experimental
2.1. Crystallization
Purified recombinant HI was provided by Novo Nordisk and crystallization was performed using the salting-out method in batch. A stock protein solution was prepared by adding 401.2 mg as-received freeze-dried insulin to 21 ml double-distilled H2O along with 2.5 ml 0.01 M zinc acetate solution, resulting in a protein concentration of 17.07 mg ml−1. Two series of samples were prepared, in which we followed the same procedure, and each one was measured in a separate diffraction experiment. Each crystallization series was performed in a pH range of approximately 4.00–8.50, keeping the other parameters constant, in order to investigate the effect of pH on the crystal forms obtained. The crystallization series are denoted Series 1 and Series 2 throughout this article.
For preparation of the protein mixture, 17 ml of the protein solution was extracted and placed in a Falcon tube along with 0.510 ml 2 M 4-ethylresorcinol diluted in DMSO. Finally, after 5 min of incubation, 0.2 ml 1 M sodium thiocyanate was added to the protein mixture. Furthermore, we prepared stock buffers of 2 M Na2HPO4 and 2 M KH2PO4. These solutions were mixed in order to produce a pH buffer range in the region of interest between 4.5 and 8.7 with steps of roughly 0.5 units for Series 1 and ∼0.3 units for Series 2. Each sample was produced by mixing 1 ml protein mixture with 250 µl pH-buffer mixture in an Eppendorf tube, giving a final protein concentration of 13.11 mg ml−1. The final concentration of 4-ethylresorcinol in each sample was 46 mM.
The samples were left to crystallize in an incubator at 298 K. After ∼48 h, polycrystalline precipitates appeared (Fig. 1). The pH of the crystallization solutions was measured before crystallization as well as after the diffraction experiments and a very small shift (∼0.2) towards higher pH levels was observed for the majority of the samples. The reported pH values in this paper correspond to the mean values of the above measurements.
2.2. Data collection and processing
High-throughput crystal screening was performed via the collection of XRPD data in our laboratory using an X'Pert PRO diffractometer (PANalytical) at room temperature (RT) (λ = 1.541874 Å). During data collection employing the laboratory diffractometer, no significant radiation damage was observed after 24 h of measurements.
High-resolution XRPD data were collected (Fig. 2) on beamline ID31 at the European Synchrotron Radiation Facility (ESRF) in Grenoble (Fitch, 2004). After loading, the capillaries were centrifuged in order to enhance the crystal packing. The capillary tubes were also spun during the measurements to avoid effects. Each sample was measured at several positions in order to counterbalance the radiation damage caused by the intense synchrotron beam, and several scans were collected at each position. Thus, the samples were translated by 2 mm every 4 min, exposing a fresh region of protein powder. The first scans at each position were combined in order to improve the counting statistics. The subsequent scans were only used in order to follow the evolution of the unit-cell parameters under exposure to X-ray radiation.
Additional measurements were performed on the materials science beamline MS-X04SA (Fig. 3) at the Swiss Light Source (SLS) in Villigen (Willmott et al., 2013). The samples were measured at RT using a wavelength of 1.37807 (15) Å and a position-sensitive Mythen detector. Each sample was loaded into a borosilicate glass capillary tube of 80 mm in length and 0.5 mm in diameter. Each sample was measured at several positions, with each scan lasting 2 s. Many scans were also collected per position, giving a total exposure time of 70 s. All data-collection parameters are listed in Table 1.
‡Modular (each module covering ∼5° in 2θ) one-dimensional solid-state Si microstrip detector (Mythen II) covering 120° in 2θ. |
The diffraction patterns were typically indexed using DASH (David et al., 2006) employing the fitted positions of at least the first 20 reflections of the high angular resolution powder diffraction profiles. From the extracted data, we were able to determine the symmetry and unit-cell parameters for all samples. In order to obtain accurate values of the unit-cell parameters and characterize the peak shape and background coefficients without a structural model, Pawley fits (Pawley, 1981) were performed using PRODD (Wright, 2004).
2.2.1. Cluster analysis
Analysis of the large amounts of data from the high-throughput screening of protein–ligand complexes (Blundell et al., 2002) is rather time-consuming without an automatic process and this is where plays a critical role (Bruno et al., 2014). The combination of XRPD methods and multivariate analysis, such as principal-component analysis, provides a rapid and effective tool for studying the influence of ligands and pH on the crystallization process (Norrman et al., 2006).
In order to investigate the crystalline properties of our protein samples throughout the pH range of interest, HighScore Plus (Degen et al., 2014). Hierarchical produced four different groups, each corresponding to one of the different crystalline phases observed in our experiments, and also indicated the most representative samples for each cluster (marked with `***' in Fig. 4). The for the samples from crystallization Series 1 is presented in Fig. 4.
was employed for the high-resolution synchrotron data sets as well as the laboratory data sets. This industrial analytical approach was performed using3. Results
The four distinct monoclinic polymorphs with the R6 identified from XRPD are now considered.
3.1. New monoclinic (P21) polymorph γ
In the case where HI was crystallized in the pH range 5.00–5.64 (Series 1) or 4.95–5.57 (Series 2), a new polymorph with monoclinic symmetry [referred to as P21(γ), unit-cell parameters a = 87.1323 (8), b = 70.294 (2), c = 48.064 (2) Å, β = 106.1729 (8)°] was observed. The same crystalline phase had been previously identified in our laboratory when insulin was crystallized with meta-cresol (m-cresol) or 4-nitrophenol (Valmas et al., 2015). The Pawley fit to the data was satisfactory, with agreement factors of χ2 = 1.0324 and Rwp = 12.427% (Fig. 5). The data acquired for the P21(γ) polycrystalline samples extended to a resolution of ∼6.5 Å.
According to Matthews coefficient calculations (Matthews, 1968; Kantardjieff & Rupp, 2003), this phase contains 12 molecules (two hexamers) per and 24 molecules (four hexamers) per corresponding to a calculated solvent content of ∼39.3% (Matthews coefficient of 2.03 Å3 Da−1). The evolution of the normalized unit-cell parameters for Series 2 of samples measured at synchrotron sources is shown in Fig. 6, while the evolution of the unit-cell volume and β monoclinic angle are shown in Fig. 7. The complete structural model of the new P21(γ) polymorph has been determined by combining traditional single-crystal and emerging analytical XRPD methods and will be presented in a forthcoming publication (Karavassili et al., 2015).
3.2. New monoclinic (P21) polymorph α
Two HI samples crystallized at pH 5.64 and 5.80 adopted a previously unknown monoclinic phase [referred to as P21(α), unit-cell parameters a = 114.130 (7), b = 336.086 (3), c = 48.987 (5) Å, β = 101.935 (8)°]. This polymorph had previously been identified by our team in the case where HI was crystallized in the presence of phenol or resorcinol, but its complete structural model remained unresolved (Karavassili et al., 2012). In this particular crystalline phase the a and b axes are considerably larger in comparison to the c axis; thus, all of the low-angle reflections used for indexing are of the form (hk0), so there is insufficient information to index a three-dimensional lattice (dominant-zone effect). Furthermore, the b:a ratio, which is very close to 3, complicates the indexing process even more since reflections such as (200) and (060) are very close in terms of 2θ and may not be observed as a result of peak overlap. In order to overcome this challenge, combined high-resolution data from the ID31 instrument and area-detector data collected at ID11 were used, as described in our previous study (Karavassili et al., 2012). The Pawley fit to the data was satisfactory, with agreement factors of χ2 = 2.5038 and Rwp = 7.095% (Fig. 8). However, the data only extended to a resolution of ∼12 Å, which does not allow structure solution and refinement.
The unit-cell volume difference in the transition from P21(γ) to P21(α) corresponds to an increase of about 6.5-fold. This is a large unit-cell modification. This crystalline phase corresponds to one of the largest HI polymorphs that we have identified via XRPD methods to date, with the other one belonging to the C2221 symmetry {V(C2221) = 3 054 394 (63) Å3, V[P21(α)] = 1 836 620 (73) Å3; Norrman & Schluckebier, 2007; Karavassili et al., 2012}. The evolution of the normalized unit-cell parameters for the P21(α) samples is shown in Fig. 9.
3.3. Monoclinic (C2) polymorph
HI crystallized in the pH range 5.97–6.23 (Series 1) or 5.88–6.21 (Series 2) adopted monoclinic symmetry [space group C2, unit-cell parameters a = 103.082 (7), b = 61.6636 (2), c = 63.5006 (4) Å, β = 117.417 (5)°]. The structure of this polymorph was first identified via XRPD methods (Norrman et al., 2006) and was subsequently fully characterized via single-crystal methods (Norrman & Schluckebier, 2007; PDB entry 2olz). The Pawley fit to the data was very good, with χ2 = 1.0619 and Rwp = 4.728% (Fig. 10). The transition from P21(α) to C2 (Series 1) resulted in a vast reduction in the unit-cell volume, which was calculated to be about five times smaller, while the transition from P21(γ) to C2 (Series 2) was calculated as ΔV[P21(γ)→C2]/ΔV[P21(γ)] = +26.9%.
According to Matthews coefficient calculations (Matthews coefficient of 2.57 Å3 Da−1) this phase contains six molecules (one hexamer) per and 24 molecules (four hexamers) per while the solvent content is ∼52.11%. The evolution of the normalized unit-cell parameters for Series 1 of crystalline samples is shown in Fig. 11. The data acquired for the C2 polycrystalline samples extended to a resolution of ∼7 Å.
3.4. Monoclinic (P21) polymorph β
In the pH range 6.73–7.94 (Series 1) or 6.70–8.10 (Series 2) the crystals obtained belonged to the monoclinic symmetry P21 [referred to as P21(β), unit-cell parameters a = 62.8231 (7), b = 62.1078 (5), c = 47.8362 (6) Å, β = 111.6913 (9)°], a previously characterized polymorph (Derewenda et al., 1989; Smith et al., 2000; PDB entry 1evr). The Pawley fit to the data was satisfactory, with χ2 = 1.8545 and Rwp = 7.926% (Fig. 12). The difference in the unit-cell volume during the transition from C2 to P21(β) corresponds to a reduction by a factor of ∼2. This polymorph is characterized by a smaller crystallographic unit-cell volume in comparison to the rest of the identified polymorphs. According to Matthews coefficient calculations, this polymorph contains six molecules (one hexamer) per and 12 molecules (two hexamers) per while the solvent content is ∼50.32% (Matthews coefficient of 2.48 Å3 Da−1). The evolution of the normalized unit-cell parameters for the two series of samples measured are shown in Fig. 13. The data acquired for the C2 polycrystalline samples extended to a resolution of ∼6 Å.
4. Discussion
In this work, we present a systematic crystallographic study of HI cocrystallized with the organic ligand 4-ethylresorcinol within the pH range 4.5–8.2. Crystallization experiments were reproduced twice and resulted in polycrystalline precipitates, which were employed for XRPD measurements. Data analysis resulted in an accurate mapping of the symmetry and unit-cell parameters for all observed distinct monoclinic crystalline phases. Four different polymorphs were identified, which belonged to two different space groups (P21 and C2). Two of these polymorphs [C2 and P21(β)] were structurally known (Fig. 14) and two were first reported by our team in previous studies [P21(α) and P21(γ)]. To date, only their unit-cell parameters and space groups have been reported (Karavassili et al., 2012; Valmas et al., 2015); complete structural models are not yet available in the PDB.
The monoclinic polymorphs observed in this study belong to the R6 (Norrman & Schluckebier, 2007; Smith et al., 2000). Insulin ligand binding clearly exhibits an allosteric behaviour (Bentley et al., 1976; Ciszak & Smith, 1994; Smith et al., 1996; Whittingham et al., 1995; Derewenda et al., 1989; Smith & Dodson, 1992). The transitions between the extended (T) and α-helical (R) conformations involve three states. In vivo, the insulin hexamer exists in the T6 state, whereas the addition of anions (for example chloride or thiocyanate) and phenolic ligands induces the T3R3 and R6 states (Dunn, 2005; Huus et al., 2006). The T-to-R transitions require the transformation of residues B1–B8 from an extended to a helical conformation (Smith et al., 1984; Derewenda et al., 1989; Ferrari et al., 2001). This transformation creates hydrophobic pockets in which phenol and its derivatives bind (Huus et al., 2006). There are three hydrophobic pockets in the T3R3 hexamer and six in the R6 hexamer (Dunn, 2005). The binding interactions of ligands in the phenolic pockets result in the T3R3 and R6 conformations, which are further stabilized by the binding of certain anions that do not stabilize the T6 state, such as halides, pseudo-halides and organic carboxylates (Bentley et al., 1976; Dunn, 2005; Rahuel-Clermont et al., 1997; Huus et al., 2006). Therefore, as most pharmaceutical preparations contain phenolic derivatives as preservatives, the HI molecules have the T3R3 and R6 conformations (Rahuel-Clermont et al., 1997; Ferrari et al., 2001). As the stability level decreases from R to T, with R6 being the most stable conformation (Rahuel-Clermont et al., 1997), the existence of these conformations may function in vivo to create a balance between the stable storage and the gradual release of the active monomer. Moreover, the allosteric transition at the level of the monomer could be important in the binding affinity between insulin and its receptor (Bloom et al., 1995).
With regard to drug development, the crystal size needs to be much smaller than 50 µm in order for pharmaceutical formulations to be easily injectable and to limit possible immunogenicity reactions (Basu et al., 2004). Control of the crystal size can be achieved by different methods such as variation of the concentration of the crystallizing agents (Collings et al., 2010) or polymeric coating, which inhibits crystal growth and reduces the particle size (Rabinow, 2004). Finally, pharmaceutical formulations need to contain isogranular and isometric crystals, which means that there should be in crystal size (Collings et al., 2010). After years of exponential development in terms of instrumentation and experiment design, XRPD is now efficient at the rapid and accurate characterization of numerous protein microcrystalline precipitates of such size with regard to and purity, whereas the extraction of accurate unit-cell parameters, as reported by our team in the present and previous studies, can indicate minor or major structural modifications.
In the case of injected treatments, it would be desirable to develop more effective formulations. This could be achieved by reducing the crystal dissolution rate and increasing the amount of active ingredient per dose. pH variation can result in distinct polymorphs with different physicochemical properties such as density, solubility and stability (Rabinow, 2004). These characteristics can further affect their dissolution rate and thus their bioavailability. Ultralente (Eli Lilly), which has been one of the most essential insulin compounds, is a suspension of insulin microcrystals that dissolve slowly following subcutaneous injection (Wagner et al., 2009). Therefore, the identification of new polymorphs could lead to the optimization of existing formulations or the design of advanced ones with a different action depending on the needs of patients. It could also lead to the creation of new forms that are associated with alternative methods of administration, such as formulations with sustained release or formulations for inhaled administration (Basu et al., 2004).
In this study, we report the identification of two new HI crystalline phases, P21(γ) and P21(α), at low pH levels (∼4.8–5.8). The P21(α) polymorph has previously been identified by our team in the presence of the ligands phenol and resorcinol in approximately the same pH range (Karavassili et al., 2012). This phase was observed only in one of the two crystallization series (Series 1, as illustrated in Fig. 15). This discrepancy may be related to seeding and the kinetics of phase formation. Moreover, the two sample series are characterized by different time intervals between crystallization and synchrotron data collection. We note that related effects have been observed in the past for other HI complexes. In addition, the P21(γ) polymorph has been identified in the cases of four different ligands, two of them recently presented (Valmas et al., 2015). The specific ligands are 4-bromoresorcinol, 4-chororesorcinol, m-cresol and 4-nitrophenol. These two polymorphs appear in an acidic environment, while the previously known polymorph P21(β) appears in the pH range ∼6.8–8.1 (Fig. 15).
The results described in the present article illustrate the extent of β angle over the pH range is shown in Fig. 16. It is obvious that the crystalline phase P21(α) has a much larger unit-cell volume owing to the long b axis.
of HI. The change in the volume andThe new P21(γ) polymorph has a unit-cell volume that is larger by 38% in comparison to the known P21(β) polymorph. The extended area is occupied by two additional hexamers [a total of four hexamers in the and 39.3% solvent content, while in the P21(β) polymorph there are two hexamers per and 50.32% solvent content]. Thus, the polymorph is characterized by a more dense packing of the insulin hexamers with much stronger intermolecular contacts. This characteristic is important for the future production of microcrystalline insulin drugs because it confers the advantage of a higher mass per volume loading, which is crucial when higher dosing is required (Rabinow, 2004). This polymorph could provide the possibility of supplying a larger amount of insulin with smaller drug doses and thereby reducing the frequency of dosing in people with diabetes.
Our results demonstrate that systematic screening of crystallization conditions in combination with synchrotron and laboratory XRPD yields an exact and unambiguous picture of the crystallization behaviour of insulin. Even around its pI (∼5.9), where its solubility is lowest and the growth of macroscopic crystals suitable for single-crystal X-ray ).
is less likely to succeed, we observed sufficient numbers of crystals, although these were smaller than those obtained at lower and higher pH values (Fig. 1Further crystallization experiments are currently in progress in order to improve the resolution of the collected XRPD profiles. We aim to obtain single crystals in order to solve and refine the structures of the new as well as the known polymorphs. This is necessary in order to identify the protein structure in each polymorph as well as the ligand-binding sites. We believe that this kind of systematic approach further extends the applicability of XRPD methods for macromolecular crystal screening in a wide range of crystallization conditions.
Acknowledgements
We would like to thank the ESRF for providing beamtime at the ID31 beamline. We would also like to thank the SLS for providing beamtime at the MS-X04SA beamline and Nicola Casati and Antonio Cervellino for assistance with data collection and useful advice, as well as PANalytical for instrumentation and software support. We are grateful to Novo Nordisk A/S for providing human insulin and for their valuable advice on crystallization procedures. IM is grateful to the UNESCO L'Oreal foundations for the award of the International Fellowship for Women in Life Sciences (2010–2012). This research has been co-financed by the following grants: the European Union (European Social Fund) in collaboration with the Greek state under the `ARISTEIA II' Action (MIS Code 4659) of the `Operational Program Education And Lifelong Learning', the European Union (European Regional Development Fund – ERDF) and Greek national funds through the Operational Program `Regional Operational Programme' of the National Strategic Reference Framework (NSRF), Research Funding Program: Support for Research, Technology and Innovation Actions in Region of Western Greece (Karatheodoris Foundation), the EU FP7 REGPOT CT-2011-285950 `SEEDRUG' project, the International Atomic Energy Agency (CRP code F12024) and the COST Action (CM1306). Finally, the company NanoMEGAS supported this work.
References
Adams, M. J., Blundell, T. L., Dodson, E. J., Dodson, G. G., Vijayan, M., Baker, E. N., Harding, M. M., Hodgkin, D. C., Rimmer, B. & Sheat, S. (1969). Nature (London), 224, 491–495. CrossRef CAS Web of Science Google Scholar
Basu, S. K., Govardhan, C. P., Jung, C. W. & Margolin, A. L. (2004). Expert Opin. Biol. Ther. 4, 301–317. CrossRef PubMed CAS Google Scholar
Bentley, G., Dodson, E., Dodson, G., Hodgkin, D. & Mercola, D. (1976). Nature (London), 261, 166–168. CrossRef PubMed CAS Web of Science Google Scholar
Bhatnagar, S., Srivastava, D., Jayadev, M. S. K. & Dubey, A. K. (2006). Prog. Biophys. Mol. Biol. 91, 199–228. CrossRef PubMed CAS Google Scholar
Bloom, C. R., Choi, W. E., Brzovic, P. S., Ha, J. J., Huang, S.-T., Kaarsholm, N. C. & Dunn, M. F. (1995). J. Mol. Biol. 245, 324–330. CrossRef CAS PubMed Web of Science Google Scholar
Blundell, T. L., Jhoti, H. & Abell, C. (2002). Nat. Rev. Drug. Discov. 1, 45–54. Web of Science CrossRef PubMed CAS Google Scholar
Brange, J. (1997). Diabetologia, 40, S48–S53. CrossRef CAS PubMed Web of Science Google Scholar
Brange, J., Langkjaer, L., Havelund, S. & Vølund, A. (1992). Pharm. Res. 9, 715–726. CrossRef PubMed CAS Web of Science Google Scholar
Bruno, A., Ruby, A., Luft, J., Grant, T. D., Seetharaman, J., Montelione, G., Hunt, J. & Snell, E. (2014). PLoS One, 9, e100782. Web of Science CrossRef PubMed Google Scholar
Carulli, L., Rondinella, S., Lombardini, S., Canedi, I., Loria, P. & Carulli, N. (2005). Aliment. Pharmacol. Ther. 22, Suppl. 2, 16–19. Google Scholar
Chang, S.-G., Choi, K.-D., Jang, S.-H. & Shin, H.-C. (2003). Mol. Cells, 16, 323–330. Web of Science PubMed CAS Google Scholar
Ciszak, E. & Smith, G. D. (1994). Biochemistry, 33, 1512–1517. CrossRef CAS PubMed Web of Science Google Scholar
Collings, I., Watier, Y., Giffard, M., Dagogo, S., Kahn, R., Bonneté, F., Wright, J. P., Fitch, A. N. & Margiolaki, I. (2010). Acta Cryst. D66, 539–548. Web of Science CrossRef CAS IUCr Journals Google Scholar
David, W. I. F., Shankland, K., van de Streek, J., Pidcock, E., Motherwell, W. D. S. & Cole, J. C. (2006). J. Appl. Cryst. 39, 910–915. Web of Science CSD CrossRef CAS IUCr Journals Google Scholar
DeFelippis, M. R., Chance, R. E. & Frank, B. H. (2001). Crit. Rev. Ther. Drug Carrier Syst. 18, 201–264. CrossRef PubMed CAS Google Scholar
Degen, T., Sadki, M., Bron, E., König, U. & Nénert, G. (2014). Powder Diffr. 29, S13–S18. Web of Science CrossRef CAS Google Scholar
Derewenda, U., Derewenda, Z., Dodson, E. J., Dodson, G. G., Reynolds, C. D., Smith, G. D., Sparks, C. & Swenson, D. (1989). Nature (London), 338, 594–596. CrossRef CAS PubMed Web of Science Google Scholar
Dunn, M. F. (2005). Biometals, 18, 295–303. Web of Science CrossRef PubMed CAS Google Scholar
Emami-Riedmaier, A., Schaeffeler, E., Nies, A. T., Mörike, K. & Schwab, M. (2015). J. Intern. Med. 277, 235–247. Web of Science CAS PubMed Google Scholar
Farr, R., Perryman, A. & Samudzi, C. (1998). J. Cryst. Growth, 183, 653–668. Web of Science CrossRef CAS Google Scholar
Ferrari, D., Diers, J. R., Bocian, D. F., Kaarsholm, N. C. & Dunn, M. F. (2001). Biopolymers, 62, 249–260. Web of Science CrossRef PubMed CAS Google Scholar
Fitch, A. N. (2004). J. Res. Natl Inst. Stand. Technol. 109, 133–142. Web of Science CrossRef CAS Google Scholar
Frankaer, C. G., Knudsen, M. V., Norén, K., Nazarenko, E., Ståhl, K. & Harris, P. (2012). Acta Cryst. D68, 1259–1271. Web of Science CrossRef CAS IUCr Journals Google Scholar
Halban, P. A., Mutkoski, R., Dodson, G. & Orci, L. (1987). Diabetologia, 30, 348–353. CrossRef CAS PubMed Web of Science Google Scholar
Hodgkin, D. C. (1971). Pure Appl. Chem. 26, 375–384. CrossRef CAS PubMed Google Scholar
Huus, K., Havelund, S., Olsen, H. B., Sigurskjold, B. W., van de Weert, M. & Frokjaer, S. (2006). Biochemistry, 45, 4014–4024. Web of Science CrossRef PubMed CAS Google Scholar
Kantardjieff, K. A. & Rupp, B. (2003). Protein Sci. 12, 1865–1871. Web of Science CrossRef PubMed CAS Google Scholar
Kantardjieff, K. A. & Rupp, B. (2004). Bioinformatics, 20, 2162–2168. Web of Science CrossRef PubMed CAS Google Scholar
Karavassili, F., Giannopoulou, A. E., Kotsiliti, E., Knight, L., Norrman, M., Schluckebier, G., Drube, L., Fitch, A. N., Wright, J. P. & Margiolaki, I. (2012). Acta Cryst. D68, 1632–1641. Web of Science CrossRef CAS IUCr Journals Google Scholar
Karavassili, F. & Margiolaki, I. (2015). Protein Pept. Lett. In the press. Google Scholar
Karavassili, F., Valmas, A., Fili, S., Norrman, M., Schluckebier, G., Wright, J., Fitch, A. & Margiolaki, I. (2015). In preparation. Google Scholar
Margiolaki, I. (2016). International Tables for Crystallography, Vol. H. In the press. Google Scholar
Margiolaki, I., Giannopoulou, A. E., Wright, J. P., Knight, L., Norrman, M., Schluckebier, G., Fitch, A. N. & Von Dreele, R. B. (2013). Acta Cryst. D69, 978–990. Web of Science CrossRef IUCr Journals Google Scholar
Margiolaki, I. & Wright, J. P. (2008). Acta Cryst. A64, 169–180. Web of Science CrossRef CAS IUCr Journals Google Scholar
Matthews, B. W. (1968). J. Mol. Biol. 33, 491–497. CrossRef CAS PubMed Web of Science Google Scholar
McPherson, A. (1985). Methods Enzymol. 114, 125–127. CrossRef CAS PubMed Web of Science Google Scholar
McPherson, A. (1995). J. Appl. Cryst. 28, 362–365. CrossRef CAS Web of Science IUCr Journals Google Scholar
Mogensen, C. E. & Zimmet, P. (2002). Diab. Metab. Res. Rev. 18, Suppl. 3, S1–S2. Google Scholar
Norrman, M. (2007). PhD thesis. Lund University, Sweden. Google Scholar
Norrman, M. & Schluckebier, G. (2007). BMC Struct. Biol. 7, 83. Google Scholar
Norrman, M., Ståhl, K., Schluckebier, G. & Al-Karadaghi, S. (2006). J. Appl. Cryst. 39, 391–400. Web of Science CrossRef CAS IUCr Journals Google Scholar
Pawley, G. S. (1981). J. Appl. Cryst. 14, 357–361. CrossRef CAS Web of Science IUCr Journals Google Scholar
Rabinow, B. E. (2004). Nat. Rev. Drug Discov. 3, 785–796. Web of Science CrossRef PubMed CAS Google Scholar
Rahuel-Clermont, S., French, C. A., Kaarsholm, N. C. & Dunn, M. F. (1997). Biochemistry, 36, 5837–5845. CAS PubMed Web of Science Google Scholar
Ryle, A., Sanger, F., Smith, F. & Kitai, R. (1955). Biochem. J. 60, 541–556. CrossRef PubMed CAS Web of Science Google Scholar
Smith, G. D., Ciszak, E., Magrum, L. A., Pangborn, W. A. & Blessing, R. H. (2000). Acta Cryst. D56, 1541–1548. Web of Science CrossRef CAS IUCr Journals Google Scholar
Smith, G. D., Ciszak, E. & Pangborn, W. A. (1996). Protein Sci. 5, 1502–1511. CrossRef CAS PubMed Web of Science Google Scholar
Smith, G. D. & Dodson, G. G. (1992). Proteins, 14, 401–408. CrossRef PubMed CAS Web of Science Google Scholar
Smith, G. D., Swenson, D. C., Dodson, E. J. & Reynolds, C. D. (1984). Proc. Natl Acad. Sci. USA, 81, 7093–7097. CrossRef CAS PubMed Web of Science Google Scholar
Valmas, A., Magiouf, K., Fili, S., Norrman, M., Schluckebier, G., Beckers, D., Degen, T., Wright, J., Fitch, A., Gozzo, F., Giannopoulou, A. E., Karavassili, F. & Margiolaki, I. (2015). Acta Cryst. D71, 819–828. Web of Science CrossRef IUCr Journals Google Scholar
Von Dreele, R. B., Stephens, P. W., Smith, G. D. & Blessing, R. H. (2000). Acta Cryst. D56, 1549–1553. Web of Science CrossRef CAS IUCr Journals Google Scholar
Wagner, A., Diez, J., Schulze-Briese, C. & Schluckebier, G. (2009). Proteins, 74, 1018–1027. Web of Science CrossRef PubMed CAS Google Scholar
Whittingham, J. L., Chaudhuri, S., Dodson, E. J., Moody, P. C. E. & Dodson, G. G. (1995). Biochemistry, 34, 15553–15563. CrossRef CAS PubMed Web of Science Google Scholar
Whittingham, J. L., Edwards, D. J., Antson, A. A., Clarkson, J. M. & Dodson, C. G. (1998). Biochemistry, 37, 11516–11523. Web of Science CrossRef CAS PubMed Google Scholar
Willmott, P. R. et al. (2013). J. Synchrotron Rad. 20, 667–682. Web of Science CrossRef CAS IUCr Journals Google Scholar
Wright, J. P. (2004). Z. Kristallogr. 219, 791–802. Web of Science CrossRef CAS Google Scholar
Zimmet, P., Alberti, K. G. & Shaw, J. (2001). Nature (London), 414, 782–787. Web of Science CrossRef PubMed CAS Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.