A broad selection of ligands was included in distinct crystallization experiments towards growing insulin crystal complexes, which were systematically studied under pH alterations. Polycrystalline precipitates (~50–100 μL volume each) of complexes, in the presence of zinc ions, were produced using the batch method, for almost all conditions studied.
Structural results, described in detail in the following sections, were obtained from diffraction patterns employing different instrumentation (Synchrotron sources: National Synchrotron Light Source (NSLS), MAX-lab synchrotron, European Synchrotron Radiation Facility (ESRF), Swiss Light Source (SLS) and laboratory diffractometers: RU200 (Rigaku Ltd., Tokyo, Japan), X’Pert PRO (PANalytical BV, Almelo, The Netherlands)) in order to optimize data quality with respect to angular resolution (FWHM) and d-spacing range.
2.1. First Human Insulin XRPD Studies
Following the first successful experiment with polycrystalline metmyoglobin, conducted by R.B Von Dreele [44
], which demonstrated that protein structure refinements using XRPD data are feasible, his research was further extended to insulin. Initially, microcrystalline slurry was produced as a byproduct of a single-crystal sample [59
] by grinding the crystals with mother liquor in an agate mortar [60
]. The slurry was placed in a glass capillary, and XRPD data were collected while the capillary was spun. Data collection was performed at room temperature at X3b1 beamline, at the National Synchrotron Light Source, equipped with a double Si(111) monochromator and a Ge(111) analyser.
From freshly made slurry, the diffraction pattern shown in Figure 1
a was obtained; whereas, material left for 3 days after grinding produced a distinctly different diffraction pattern as shown in Figure 1
b. The pattern from the ground material was indexed in rhombohedral symmetry, with a
= 81.9678 (7) Å, c
= 37.5914 (8) Å, identical to the single-crystal unit cell for T3
HI conformation [19
], whereas the pattern from the freshly ground material, revealed a previously unknown rhombohedral polymorph with a
= 81.2780 (7) Å, c
= 73.0389 (9) Å, which is fundamentally a doubled c axis superlattice of the T3
structure (a phase denoted as T3
Owing to the close relationship between these two phases, the structure solution of T3
DC using the molecular-replacement technique was employed. A starting model was introduced from the single-crystal coordinates for the T3
], and a three-parameter (two rotation angles and one translation) rigid-body Rietveld refinement was later performed. Atomic coordinates, extracted from stereochemically restrained Rietveld refinement of the T3
crystal structure, were used to complete the rigid-body refinement of the T3
The complete structural characterization of the T3
DC insulin form achieved via XRPD was also verified via single crystal experiments one year later [59
], and revealed a number of special features of this new variant of the T3
human insulin-Zn complex. After grinding, a reduction of the material’s volume by 2.095% or 1490 Å−3
complex was evident, which consequently induced a structural change resulting in c axis doubling of the rhombohedral unit cell. One of the independent dimers rotates 17.2° about the c axis in the conversion from T3
DC; the other rotates 9.5° in the same direction (Figure 2
). This rotation is probably associated with a collapse of the spacing between the pairs of (AB)2
complexes along the crystallographic c axis, and a repositioning of B chains with extended conformation. Conceivably, water molecules extracted from the structure during grinding could originate from this particular location.
This was one of the first research results demonstrating the applicability of powder diffraction method for macromolecular crystal screening and detailed structure solution of a protein molecule. Within the next five years, continuous developments in instrumentation as well as in data collection and analysis were carried out in parallel by Robert Von Dreele at Argonne National Laboratory (USA) and Irene Margiolaki and colleagues at ESRF (Grenoble, France). Their early studies on lysozyme (Turkey or Hen egg-white) as a model system further established the use of XRPD as a valuable tool in the identification of small structural variations in protein molecules [49
2.2. Characterization of Distinct Insulin Formulations Via XRPD
Along with the underlying difficulties of developing and producing biopharmaceutical compounds, the characterization of the final product can sometimes be even more challenging and demand a repeated revision process of analytical methods performed in a high throughput manner, without compromising the accuracy of the obtained results. On top of this, protein therapeutics correspond to a class of products which have an intricate structure whose integrity determines the bioavailability, biological activity, clinical efficacy, and safety. All factors which control the aforementioned characteristics of a product are extensively studied in the production processes, and provide valuable information for further refining the enzyme/protein manufacturing.
The first study of this kind was originally conducted in 2006 by Norrman et al. [54
], where 12 insulin formulations (some commercially available) were investigated via XRPD. Despite the medium-resolution XRPD patterns obtained, the data in combination with multivariate data analysis were used to compare insulin microcrystals preparations.
The commercially available insulin preparations examined in that project (Ultratard, Ultralente, Lente, Detemir, Penmix30, Novomix30 and Protaphan) were obtained from Novo Nordisk A/S, whereas additional microcrystals were prepared following the batch crystallization method. All products examined were “descendants” of the first stable protracted insulin formulation, the Neutral Protamine Hagedorn (NPH), which was introduced in 1946 [64
]. This formulation was based in an observation by Hans Christian Hagedorn (founder of former Novo Nordisk A/S) and B. Norman Jensen in 1936, introducing that the effects of injected insulin could be prolonged with the addition of protamine—a peptide consisting mainly of arginine—obtained from the semen of river trout. An insulin–zinc solution was cocrystallized with protamine, reducing insulin’s solubility and resulting in NPH insulin; an intermediate–acting insulin product.
Among all HI crystals produced by batch crystallization, two novel crystal types were obtained. Orthorhombic C
= 59 Å, b
= 219 Å, c
= 223 Å) with three hexamers in the asymmetric unit, adopting the R6
configuration were identified in presence of urea, NaCl, and resorcinol at pH 6.7 [54
], whereas in slightly higher pH values (~7) monoclinic C
2 crystals (a
= 100 Å, b
= 60 Å, c
= 62 Å, β
= 116°) were observed containing one hexamer with R6
molecular conformation in the asymmetric unit, and 50% solvent content [55
]. Crystallization conditions for all formulations used in that study are summarized in Table 1
Protein powder data of this study were collected at room temperature, both in-house (on a Mar345 imaging plate detector, using an RU200 rotating anode, λ = 1.5418 Å, Rigaku Ltd.) and at the MAX-lab synchrotron (Lund, Sweden), beamlines 711, 911-2 and 911-3 [66
], using charge-coupled device (CCD) detectors. Data indexing was in all cases unsuccessful, even though a variety of software was exploited, due to low angular resolution (broad overlapping diffraction peaks) and the use of area detectors, which resulted in further peak overlap. Thus, only synchrotron powder diffraction patterns were employed for extracting preliminary structural information, due to their advantageous d-spacing and angular resolution. Nevertheless medium-resolution powder diffraction patterns were enough for effective classification in crystal systems via Principal Component Analysis (PCA) [68
]. Crystallographic properties of all samples described in this project are listed in Table 2
Patterns from different insulin polymorphs showed distinct peaks in the low 2θ
region (0.9° to ~6°). Visual evaluation of the plots in Figure 3
shows that crystals, belonging to the same crystal system according to the bibliography with the same type of structure, have very similar powder patterns as well. Despite the fact that powder patterns have been collected without the optimum instrumentation, they reveal even small differences in protein structure based in alternations in peaks’ positions (F, D, and E crystals), and/or the extinction of several peaks (I, J, and K).
F, D and, E crystals all belong to the rhombohedral space group R
3 with T6
molecular conformations, respectively. As seen from Figure 3
(left panel), similar peaks in the three patterns are generally shifted by less than 0.12° in 2θ. Peak variances are more evident within the 2θ range of 3.95 to 4.35°, where in all cases a high-intensity peak is observed, but its position is clearly different. The shifts in peak positions are associated to structural differences in the N-terminal part of the B-chain, causing alternation in crystal packing and thus in the unit cell constants; especially in the length of the c-axis.
Crystals I, J, and K belong to the same space group according to Table 2
. Powder patterns from the three types of crystallites share a high degree of similarity, especially in the low 2θ region, as shown in Figure 3
(right panel). The major difference among them is an additional peak at 2θ = 4.1° in the K pattern (marked with an arrow in Figure 3
c) that is not found in the I pattern. Also, peak positions in the J pattern are shifted relative to I and K patterns in the entire region, reflecting the slightly larger unit-cell parameters of J crystals (Table 1
). This can be explained considering the mutation B28Asp in J crystals, which alters the molecule’s charge, thus a higher proportion of the protamine peptide is being bound on insulin [65
], resulting in slightly altered unit-cell parameters.
Visual analysis of the powder patterns described above, demonstrates that even without successful data indexing, the method can be used to effectively distinguish different crystal systems and assess homogeneity of different batches or preparations of insulin. However, the complexity increases when examining a plethora of microcrystal suspensions, and the procedure can be time-consuming, thus Norrman et al. [54
], employed the PCA analysis to facilitate the interpretation of powder patterns. Through PCA, data dimensionality (number of variables) is reduced, via a statistical procedure, from several hundreds to two or three principal components, resulting in a visual representation of the relationships and similarities of the—powder patterns of the—samples, by grouping them into clusters. Diffraction patterns from the crystals mentioned above were represented as data points, and their clustering indicated a high similarity feature within each group. For example, the relative shifts in peak positions of the three rhombohedral D, E, and F crystals, due to distinct B-chain conformations (R6
respectively) had a large impact on the distribution of their PCA scores in the plot, and thus were not clustered together. Following this approach, different crystal systems and/or structural arrangements can be clearly separated, further facilitating the detection of novel polymorphs as in the case of B and X type of crystals, which were clearly distinguished from other clusters.
The identification of two novel crystal forms (orthorhombic C
and monoclinic C
2, Figure 4
) of human insulin accomplished in this project declare the use of XRPD as a powerful approach for characterization and evaluation of macromolecular microcrystalline suspensions, both during polymorph screening, and in manufacturing process control. The medium-resolution data of the early XRPD era did not allow for detailed structural characterization, thus this was achieved a year later [55
] via single crystal experiments (Protein Data Bank (PDB) codes: 2OM1 for the C
crystal form and 2OLZ for the C
2 crystal form).
The discovery of novel insulin polymorphs from Norrman & Schluckebier [55
] triggered the research around insulin, and variations in cocrystallization and pH conditions forced the discovery of several other insulin crystalline polymorphs waiting to be examined in terms of physical stability, dissolution rate, and other bioavailability properties.
Bovine insulin polycrystalline precipitates were extensively studied later on as well, in a wide pH range 5.0–7.6. Powder X-ray diffraction data revealed the T6
hexameric insulin form (space group R
3; unit-cell parameters a
= 82.5951 (9) Å, c
= 33.6089 (3) Å for the sample crystallized at pH 5.0) in agreement with the high-resolution structure of HI, identified earlier by single crystal experiments [69
Fourteen powder diffraction profiles with slightly different lattice parameters were selected for structure analysis. Lattice parameters variations were caused by alterations in the sample preparation procedure, or were induced by radiation exposure. In the diffraction patterns, these variations are depicted by shifts in the positions of adjacent peaks, allowing the contributing reflections of the overlapped peaks to be partially deconvoluted. Stereochemically restrained Rietveld refinement was performed to obtain an average crystal structure of bovine insulin over the pH range using the General Structural Analysis System (GSAS) software [71
Selected regions of the refined coordinates and the total OMIT map [73
] computed at the final steps of analysis are presented in Figure 5
. Each of the two zinc ions in the hexameric structure is octahedrally coordinated by three Nε2
atoms of three symmetry-related HisB10 residues and three symmetry related water molecules (PDB code: 4IDW).
The successful identification of the above formulations has reinforced the use of powder diffraction, by our group, as a rudimentary tool in daily research, for investigating the structural behavior of HI in a wide range of crystallization conditions in terms of pH and addition of ligands.
2.3. Cocrystallization of HI with Phenolic Derivatives and pH Dependence
Phenol and phenol-like compounds have been added in insulin formulations as antibacterial agents since the earliest years of production. It is well known that phenol binds in pockets of the insulin hexamer and alters intensively insulin’s conformation, driving it to the R state [22
While varying the pH in the presence of phenolic derivatives, a series of phase transitions has been reported. Specifically in the case of cocrystallization with phenol, four distinct polymorphs have been identified, three polymorphs with resorcinol, two with m
-cresol, and 4-nitrophenol and six with 4-ethylresorcinol (Table 3
The quality of the obtained data allowed for successful indexing, using the fitted positions of at least 20 first reflections of each diffraction profile. From the extracted data, symmetry and unit-cell parameters were effectively determined.
When HI was crystallized with phenol, in addition to the earlier identified polymorphs C
], and P
], a new monoclinic phase of insulin has been detected (Figure 6
) within the pH range 5.47–5.70, space group P
, (referred thereafter as P
). Indexing of this unit cell was particularly challenging due to dominant-zone problem, as the majority of low two-theta reflections belong to the dominant zone in reciprocal space. These reflections initially were not detected owing to peak overlap, however, combined use of diffraction data collected with different detectors confirmed the existence of a screw axis, and led to the identification of the monoclinic cell P
with remarkably large unit-cell parameters a
= 114.682 (6) Å, b
= 337.63 (2) Å, c
= 49.270 (4) Å, β
= 101.555 (6)°, which originally caused the dominant zone effect. Diffraction profiles acquired from P
crystals extended to ~7.5 Å resolution. This was the first report of this specific crystallographic phase of human insulin.
HI exhibited similar behavior as with phenol, when crystallized with resorcinol at pH 5.29 and 5.46, yielding the same monoclinic phase (space group P21, unit-cell parameters a = 114.0228 (8) Å, b = 335.43 (3) Å, c = 49.211 (6) Å, β = 101.531 (8)°).
The discovery of a previously unknown crystal form of insulin was the result of a systematic study of the effect of pH—even around its isoelectric point (~5.9)—on the crystallization behavior of insulin in complex with zinc and a phenolic ligand. Nearby the pI region, its solubility is lowest and growing macroscopic crystals suitable for single-crystal X-ray structure determination is least likely to succeed. The novel insulin crystal packing, was identified in this exact pH area in the presence of phenol or resorcinol through XRPD, and that is (probably) the reason why the monoclinic P
conformation remained undetected even though crystallization experiments with phenol and resorcinol have been earlier reported [20
Nevertheless, the earlier identified insulin forms (C
2) were obtained in these studies as well. Human insulin crystallized in the presence of phenol (pH 5.93–6.54), and resorcinol (pH 5.93–7.45) produced crystals with orthorhombic symmetry (space group C
) containing three protein hexamers per asymmetric unit [55
In both cases, the pH increment caused slight lattice parameter alterations, as illustrated by the smooth anisotropic shifts in the peak positions and no indication of a first-order phase transition. Apart from the C2 phase, which was only observed during cocrystallization with phenol, all other phases obtained, coincided in crystallization experiments with the two ligands exhibiting minor alterations in unit-cell parameters.
Although phenol and resorcinol can substitute each other as allosteric ligands of the insulin hexamer without detectable changes in insulin structure [28
], the presence of ligand apparently influences the crystallization behavior. This is noteworthy, concerning that phenolic binding sites are far from the interfaces or the location of crystal contacts. Results from the systematic screening of crystallization conditions suggest that human insulin crystallized in the presence of phenol and resorcinol is greatly affected by pH. This analytical approach further extends the applicability of powder diffraction methods for efficient macromolecular crystal screening. Specifically, when synchrotron XRPD patterns are employed in the analysis, the low instrumental contribution to the diffraction peaks, resulting in accurate peak positions, allows for high precision in unit-cell parameters determination, and thus small variations of lattices can be quantified precisely.
The structural behavior of HI when cocrystallized with two widely used phenol-based ligands, m
-cresol and 4-nitrophenol was further examined in a broad pH range [56
]. These organic additives, were selected as they can serve as bactericidal agents and earlier structural results on HI complexed with these exist in the literature [27
]. Particularly m
-cresol comparing to phenol, seems to be a more effective germicide, and is widely used as an antimicrobial preservative in pharmaceutical formulations [74
Several polycrystalline samples were produced, and consecutive data collection experiments were performed using various X-ray sources to exploit their influence on diffraction patterns and to ensure the validity of the results. A thorough data analysis revealed a first order phase transition with pH variation, resulting in two distinct polymorphs in both cases (Table 3
), whereas a novel monoclinic phase of insulin was identified (space group P
, referred in the following as P
). Specifically when HI was crystallized with m
-cresol (pH range 4.50–6.70) or 4-nitrophenol (pH range 5.1–6.3), this new monoclinic polymorph was identified (Figure 7
) with the following lattice parameters, a
= 87.0749 (7) Å, b
= 70.1190 (5) Å, c
= 48.1679 (5) Å, β
= 106.7442 (8)°. The diffraction patterns obtained for the P
polycrystalline samples yield a d-spacing of approximately 6.8 Å.
While moving towards neutral or basic pH regions, a first-order transition occurs, as it is evident in Figure 8
. The monoclinic symmetry transforms into a rhombohedral symmetry (space group R
3) that is stable over a wide pH range (approximately 6.2–8.1) consisting of three protein hexamers per unit cell.
Data analysis of XRPD profiles of HI cocrystallized with 4-nitrophenol, led to the accurate extraction of the following lattice parameters a
= 80.721 (1) Å, c
= 37.8039 (5) Å, γ
= 120.000° for the sample crystallized at pH 6.41. From the parameters obtained it is derived that HI cocrystallized with this ligand acquires the T3
]. XRPD profiles collected on ID31 (now ID22) for these samples extended to a resolution of 3.6 Å.
-cresol is employed in insulin crystallization at pH 6.7–8.6, the R
3 space group is identified with slightly altered unit-cell parameters. Pawley analysis of high-resolution diffraction profiles resulted in: a
= 80.0644 (6) Å, c
= 40.8396 (3) Å, γ
= 120.000° for the sample crystallized at pH 8.15. These values indicate that HI acquires the R6
]. XRPD profiles collected on ID31 (now ID22) for these samples extended to a d-spacing resolution of 3.7 Å.
Thorough examination of the lattice parameters close to the transition from the monoclinic crystal type to the rhombohedral one, yields a decrease in the unit-cell volume of about ΔV(P21(γ)→R3)/VP21(γ) = −25.53%, while for HI complexed with m-cresol the cell is reduced by ΔV(P21(γ)→R3)/VP21(γ) = −21.89%.
Comparing to the isosymmetrical polymorph, P
, identified by Karavassili et al. in 2012 [29
], which exhibited remarkably large cell dimensions concerning a
= 114.0228 (8) Å, b
= 335.43 (3) Å, c
= 49.211 (6) Å, β
= 101.531 (8)°), the lattice parameters of this new polymorph P
are significantly shorter, approaching the already known range of dimensions that other known monoclinic cells adopt [22
]. Between these two monoclinic forms and the already deposited in the Protein Data Bank P
(PDB code: 1EV6; [28
]), an unusual crystal packing for the P
polymorph is noteworthy. While P
, consists of six molecules per asymmetric unit, and 48% solvent content, according to Matthews Coefficient calculation [75
], the novel P
polymorph contains twelve molecules per asymmetric unit and 39% solvent content (Matthews coefficient = 2.03 Å3
). This difference between the cell contents among the two polymorphs, reveals a denser crystal packing in the case of P
which could be of great pharmacological importance.
Interhexamer interactions that may form owing to the very dense packing of the polymorph could associate with enhanced physicochemical properties whereas in the case of crystalline insulin formulations this can be interpreted as increased stability, and thus provide a prolonged formulation lifetime. This could be a key point with a significant impact in the formation of new types of insulin-based microcrystalline preparations for treating diabetes. Furthermore, the preparation of pharmaceutical products consisting of crystals with high protein concentration could lead to minimization of injection times.
The complete structure determination of the novel P
polymorph has been derived from the combined use of traditional single-crystal and emerging XRPD approaches and will be presented in a forthcoming publication by our team [77
The ligand 4-ethylresorcinol, a strong antiseptic and disinfectant of pharmaceutical formulations, was used during systematic crystallization experiments of HI in the presence of zinc ions as well [53
]. Diffraction patterns obtained from several sources from crystals grown within the pH range 4.50–8.20 revealed four different crystalline polymorphs (Table 3
). Among these, the two new monoclinic symmetry phases (P
) described earlier, were detected again, emphasizing their characterization as potential targets for the future development of microcrystalline insulin drugs.
The large quantity of diffraction patterns derived in this study were initially handled via PCA using HighScore Plus software [78
], which classified patterns in four distinct groups (Figure 9
), corresponding to the mentioned crystalline phases, and indicated also the most representative sample of each cluster (marked with ***).
Systematic data analysis confirmed the three first order phase transitions with pH variation, observed in PCA analysis, which resulted in four distinct polymorphs of monoclinic symmetry (space group P
2). Accurate unit-cell parameters of each polymorph are presented in Table 3
Specifically, when HI was crystallized in the presence of 4-ethylresorcinol, within the pH range 4.95–5.80, two novel polymorphs with monoclinic symmetry (P
, in pH range 4.95–5.60 with lattice parameters a
= 87.1323 (8) Å, b
= 70.294 (2) Å, c
= 48.064 (2) Å, β
= 106.1729 (8)° and P
in pH range 5.65–5.80 with lattice parameters a
= 114.130 (7) Å, b
= 336.086 (3) Å, c
= 48.987 (5) Å, β
= 101.935 (8)°) were observed (Figure 10
). These crystalline polymorphs had been identified earlier by XRPD [29
]. Diffraction profiles acquired for the P
polycrystalline samples extended to a resolution of ~6.5 Å, whereas the lower resolution range for the P
polycrystalline samples (~112–12 Å) was sufficient for successful indexing and Pawley analysis.
According to Matthews coefficient calculations [75
], the P
phase contains 12 molecules (two hexamers) per asymmetric unit and doubled molecules per unit cell, corresponding to ~39.3% solvent content (Matthews coefficient of 2.03 Å3
). The volume of the cell while shifting from P
increases of about 6.5-fold resulting in a significant unit-cell modification. Summarizing these results in terms of cell volume, the P
is one of the largest phases that has been identified to date through XRPD, while the C
phase is being sorted as the second one (V(C2221)
= 3.054.394 (63) Å3
= 1836620 (73) Å3
]). Furthermore, from the crystallization of HI in the presence of 4-ethylresorcinol in the pH range ~6.00–8.00 the crystals obtained, belonged to monoclinic symmetry (space group C
2 (pH 5.93–6.25) and P
(pH 6.73–8.05)). The complete structural characterization of these polymorphs has been determined and thoroughly described previously (PDB code 2OLZ [55
] and PDB code 1EVR [22
The systematic crystallization experiments of HI in the presence of 4-ethylresorcinol within the pH range 4.5–8.2 resulted in a discrete characterization of the observed polymorphs in terms of crystal symmetry and lattice parameters. Insulin in these polymorphs adopts the R6 molecular conformation of B chain, where binding interactions of ligands in the phenolic pockets seem to stabilize the specific conformation; a process assisted by a number of certain anions such as halides, pseudohalides and organic carboxylates.
This conformation is commonly apparent in pharmaceutical preparations, as most of them contain phenolic derivatives as disinfectants, driving HI either to the T3
molecular conformations [79
]. Concerning that stability levels increase from T to the R state [79
], the existence of the most stable conformations in formulations can serve two principal aspects: sufficient storage stability of the pharmaceutical preparations, and gradual release of the active monomer once the formulation is injected into the human body. Moreover, the allosteric transition at the level of monomer could be proven as essential for the binding affinity of insulin to its receptor [81
These observations could be of great importance with regard to the improvement of injected preparations, as by reducing crystal’s dissolution rate and increasing the amount of active ingredient per dose would result in more effective formulations. Variations in the pH, during crystallization procedures, can induce the formation of distinct polymorphs with different physicochemical properties such as density, solubility, and stability [82
]. These characteristics can further affect the dissolution rate, and thus the bioavailability of the final pharmaceutical products. Therefore, the identification of novel crystalline polymorphs could aid towards optimizing existing formulations, or designing advanced preparations with improved action and characteristics, in accordance with patients’ needs, including preparations associated with alternative methods of administration, such as formulations with sustained release or formulations for inhaled administration [37
Several HI polymorphs described in this study are summarized in Figure 11
, with respect to the ligand and the pH values each polymorph appears.
2.5. Ligand-Free Crystalline HI Studies and pH Dependence
Towards the direction of understanding better the effect of pH upon HI conformational changes, further crystallization experiments were performed in a wide pH range (4.88–8.56) without the presence of any ligand.
Specifically, HI was crystallized using a solution of 13.14 mg/mL protein concentration, in the presence of 0.8 mM zinc acetate, 10.25 mM sodium thiocyanate, and 0.4 M sodium/monopotassium phosphate buffers of ascending pH per sample, in order to investigate the influence of pH on insulin crystallinity and conformation.
Diffraction patterns from both synchrotron and laboratory sources were collected and indexed successfully. Systematic data analysis led to the identification of 2 different space groups; R
3 rhombohedral (pH range 4.9–7.7), and I
3 cubic (pH range 7.75–8.60). Within the acidic pH, the T6
configuration (space group R
= 82.9943 (5) Å, c
= 34.0642 (2) Å) was identified while in pH range 6.9 to 7.7, the T6
alters to T3
= 80.6630 (5) Å, c
= 37.7459 (2) Å), a transition that is evidently depicted in peak’s positions changes (Figure 13
). The figure indicates an additional structural modification from samples with even higher pH values (7.8–8.6). A first order phase transition occurs at pH around 7.7, and insulin molecules obtain a cubic symmetry (space group I
= 78.9 Å, PDB code: 9INS) [30
All diffraction patterns of this study were collected on ID22 at the ESRF, and extend to a resolution of 3.3 Å (R
3 polymorphs) and 2.7 Å (I
3 polymorph) as illustrated in Figure 14
HI crystals grown in solutions with pH higher than 7.7 adopt the cubic symmetry, which is the most common zinc-free crystal form, in accordance with bibliography [92
]. However, zinc ions, mandatory for HI hexamer formation [94
], were initially added during crystallization. Consequently, we conclude that in alkaline conditions, zinc ions are not able to interact with the molecule, and for this reason HI crystals are formed from dimers and not from hexamers. The accuracy of this allegation was verified via a structure solution of a microcrystalline sample at pH 8.56 from powder diffraction data (d-spacing resolution ~2.5 Å), which clearly revealed the absence of zinc ions from their common binding sites (Figure 15
): two identical high-affinity sites located on the three-fold symmetry axis near histidines in the two distinct symmetries. Detailed description of the cubic structure will be discussed elsewhere [91
This phenomenon can be explained by considering the charge of all different ionizable groups of insulin molecule. For pH ≥ 7.5, histidines, due to imidazole rings’ acid dissociation constant pKa (7.5), are neutral [95
]. Uncharged His cannot associate with zinc ions, and consequently insulin hexamers cannot form.
This observation could be of great importance for the pharmaceutical industry. The majority of the commercially available compounds consist of crystals containing HI hexamers, tightly packed within the unit-cell, allowing a minimum amount of solvent. However, it is evident that even slight alterations in storage conditions (e.g., temperature), which can directly affect parameters such as pH, may alter the tertiary molecular structure modifying physicochemical characteristics of the molecule and drug’s ADME.