Accurate hydrogen parameters for the amino acid L-leucine

Synopsis Modern neutron Laue diffraction has been used to determine the structure of the primary amino acid L -leucine. Abstract The structure of the primary amino acid L -leucine has been determined for the first time by neutron diffraction. This was made possible by the use of modern neutron Laue diffraction to overcome the previously prohibitive effects of crystal size or quality. The packing of the structure into hydrophobic and hydrophilic layers is explained by the intermolecular interaction energies calculated using the PIXEL method. Variable-temperature data collections confirmed the absence of phase transitions between 120 and 300 K in the single-crystal form.


Introduction
The advantages of neutron diffraction for providing accurate geometric parameters for amino acids and other molecular materials are well known, and include the strong and contrasting scattering lengths of hydrogen and deuterium, and the absence of form-factor fall off with scattering angle (Niimura & Bau, 2008, McIntyre, 2014, Görbitz, 2015. Accurate geometric parameters for hydrogen atoms are especially important as structures derived from X-ray studies suffer from sever systematic errors as a result of aspherical electron density distributions about covalently-bonded hydrogen. Overall, a total of 16 of the 20 naturally-occurring amino acids have been subject to structure determination by neutron diffraction In the early 1970s Hamilton and colleagues at Brookhaven National Laboratory refined the structures, including the all-important hydrogen atoms, of 13 of the 2 20 naturally-occurring amino acids in an ambitious series of single-crystal neutron diffraction studies (see references below). Three further structures came from experiments at the Indian Atomic Energy Laboratory, Trombay, to create a library of accurate and complete structures. The derived X-H restraints need to be shortened for X-ray protein refinements due to the displacement of the single electron on the H atom, but the neutron library gives considerably more accurate bond distances than an X-ray-based library. Of course, the derived X-H restraints apply without adjustment to refinement of protein structures based on neutron data, which is growing in importance thanks to experimental improvements of the type described below (Munshi et al., 2012).
To date, a total of 16 of the 20 common amino acids have been the subject to structure determination by neutron diffraction. Of the amino acids with electrically charged side-chains structures have been obtained for L-arginine , L-histidine (Lehmann et al., 1972a), L-lysine , and L-glutamic acid (Lehmann et al., 1972b, Lehmann & Nunes, 1980 in the enantiopure form, aspartic acid was determined in the racemic DL-aspartic acid form (Sequeira et al., 1989). Of the polar uncharged amino acids neutron-diffraction-derived structures have been determined for L-serine , L-threonine (Ramanadham et al., 1973b), Lasparagine , Ramanadham et al., 1972, Weisinger-Lewin et al., 1989, and Lglutamine . Of the hydrophobic-side-chain amino acids, structures have been determined by neutron diffraction for L-alanine (Lehmann et al., 1972c, Wilson et al., 2005, L-valine , L-phenylalanine (Al-Karaghouli & Koetzle, 1975), L-tyrosine  , and L-tryptophan (Andrews et al., 1974). Of the remaining amino acids structures have been determined for L-cysteine (Ramanadham et al., 1973a), and L-glycine in both α (Jonsson & Kvick, 1972) and γ polymorphs (Kvick et al., 1980). Amino acid structural parameters derived from neutron diffraction have found extensive application as constraints and restraints in macromolecular refinements, and many of the entries above are still the preferred standards today and remain heavily cited in the literature. These publications have a mean number of citations of 83 overall, and 20 in the last five years. For example, neutron structures are used for restraints applied to C, N and O positions in the program suite PROTIN/PROLSQ (Konnert & Hendrickson, 1980), and are particularly valuable in when applied to joint refinement of X-ray and neutron data (Wlodawer & Hendrickson, 1982). They are also used for validation of refined protein structures. Hydrogen constitutes ca. 50% of the atoms in macromolecules (Myles, 2006), and it is essential to include it in refinement models. Although the derived X-H restraints need to be shortened for X-ray protein refinements, the neutron library gives considerably more accurate bond directionality than an X-ray-based library (Konnert, 1976, Wlodawer & Hendrickson, 1982, Hendrickson & Konnert, 1981, Teeter & Kossiakoff, 1983, Niimura et al., 1997.
The neutron structure of L-leucine has not been reported. The Brookhaven and Trombay studies required single crystals with volumes of 10 mm 3 or more, and for the three amino acids, L-leucine, Lisoleucine, and L-methionine, only crystals of volumes suitable for X-ray diffraction could be grown. L-leucine proved to be particularly troublesome, but methods for growing crystals of volumes of 0.1 mm 3 are now known (Görbitz & Dalhus, 1996b). The use of Laue (white beam) diffraction coupled with advances in neutron image-plate technology has increased the range of applicability of neutron crystallography (Cole et al., 2001, McIntyre et al., 2006, and the LADI and VIVALDI instruments at the ILL pioneered the application of this technique to macromolecular and small-molecule crystallography, respectively (Cipriani et al., 1996, Wilkinson et al., 2002. The technique is eminently suitable for crystals with volumes of 0.1 mm 3 (McIntyre et al., 2006, Aznavour et al., 2008, Edwards, 2011. Here we report the neutron-diffraction-derived structures of L-leucine at 120 K and room temperature as determined using the KOALA Laue diffractometer at ANSTO. The structure of L-leucine was first determined by Harding & Howieson (1976) with subsequent redeterminations by Coll et al. (1986) and most recently Görbitz & Dalhus (1996b) as part of a series of redeterminations of amino-acid structures. When studied using single-crystal methods, the roomtemperature phase of L-leucine persists to 120 K; the structure at this temperature is monoclinic, P21, a = 9.562(2), b = 5.301(1), c = 14.519(3) Å, β = 94.20(2)° (Görbitz & Dalhus, 1996b).
By contrast, when studied by power diffraction L-leucine has been reported to undergo three phase transitions at 150, 275, and 353 K (Façanha Filho et al., 2011). A combination of calorimetric and Xray powder diffraction data was used to identify two transitions at T1 = 150 K and T2 = 275 K.
Additional peaks in the X-ray powder diffraction data were taken as an indication of a doubled a unitcell length, however a limited 2θ range precluded a Rietveld analysis (Rietveld, 1969, Façanha Filho et al., 2011. The unit-cell dimensions derived from Le Bail fitting (Le Bail et al., 1988) exhibit two sharp discontinuities in the β angle. Analysis of the systematic absences was ambiguous, but the requirement for enantiopurity implied either P2 or P21. As noted by Façanha Filho et al. (2011), there is a need for additional high-resolution diffraction data to characterise the transitions fully.

Crystallisation, data collection and refinement
A single colourless plate of 2 x 0.5 x 0.2 mm 3 of L-leucine (Aldrich) was grown from a warm (ca. 323 K) concentrated aqueous solution which was allowed to cool to room temperature (Harding & Howieson, 1976, Görbitz & Dalhus, 1996b. Neutron diffraction data were collected on the KOALA quasi-Laue diffractometer, ANSTO at 120 and 300 K. Data were collected in two orientations of the crystal relative to the single (vertical) rotation axis of the instrument in order to optimise completeness. The exposure time was two hours per pattern at both temperatures.
At 120 K, nine Laue patterns were collected in each crystal setting giving a total of 18 patterns.
Patterns were related by a 20° rotation about the vertical axis in the range −90° < φ < 90°.
Short collections (10 minutes each) were also carried out during the heating and cooling process to check for signs of additional Bragg peaks that would be diagnostic of phase transitions. Cooling the sample crystal to 120 K led to the appearance of splitting in the Laue spots. The degree of splitting was not uniform over the detector surface indicting that the splitting arises from a macroscopic movement of layers making up the crystal, rather than changes in atomic structure. Furthermore subsequent heating of the sample back to room temperature led to no significant changes in the shape of the Laue diffraction spots. Structure factors could nevertheless be extracted from these patterns, although the overall quality of the data at 120 K is a little lower than for those at room temperature.
The Laue diffraction patterns were indexed and processed using the program LaueG (Piltz, 2011).
Reflection intensities were integrated with a modified two-dimensional version of the algorithm formulated by Wilkinson et al. (1988) and Prince et al. (1997). Resolution limits were determined based on the shortest d-spacing at which 5% of reflections had I/σ(I) > 5. The data were empirically normalised to a single common incident wavelength using the program Laue4 (Piltz, 2011), by comparison of repeat observations and equivalent reflections with wavelengths within the range λ = 0.80-1.7 Å; reflections outside this range were too weak or had too few repeat measurements or equivalents to be able to determine the normalisation curve with confidence. Absorption or extinction corrections were deemed unnecessary on account of the small sample size.
Refinement of the crystal structures was carried out against |F 2 | with the SHELXL refinement package (Sheldrick, 2015) using least-squares minimisation with initial atomic coordinates provided by (Görbitz & Dalhus, 1996b). As a result of the polychromatic incident beam, Laue diffraction is not capable of determining absolute values of the unit-cell dimensions using the observed reflection coordinates, only ratios of a:b:c. Unit-cell dimensions for the room temperature and 120 K data-sets were therefore taken from corresponding X-ray diffraction studies (Coll et al., 1986, Görbitz & Dalhus, 1996b. Figure 1 shows the refined wavelength spectra for data collected at 300 K and 120 K, and the nominal instrument spectrum. The shifts in wavelength distribution are negligible; −0.7% for 120 K and +0.3% for 300 K, implying that the unit-cell dimensions provided by the X-ray data of Görbitz & Dalhus (1996b) and Coll et al. (1986) match our data.

Figure 1
Refined and normalised instrument wavelength spectra for Laue data collected at 120 K and 300 K, the nominal spectrum is included for comparison.
Absolute structure Absolute structure known from synthesis Absolute structure known from synthesis Values in italics are derived from X-ray diffraction measurements. 1 (Coll et al., 1986) 2 (Görbitz & Dalhus, 1996b) Molecular geometries were analysed using PLATON (Spek, 2009). Intermolecular interaction energies were calculated using the PIXEL method (Gavezzotti, 2005(Gavezzotti, , 2011. Electron densities were 7 calculated using Gaussian09 at the MP2 level of theory with the 6-31G** basis set (Frisch et al., 2009). PIXEL calculations were accomplished with the PixelC module of the CLP package which allows the calculation of dimer and lattice energies.
Hirshfeld surfaces, which enable graphical comparison of molecular interactions for similar configurations, were calculated using CrystalExplorer 3.1 (Wolff et al., 2012). The Hirshfeld surface for a given molecule in a given crystal is an isosurface calculated from the ratio of the molecular electron density (the promolecule) over the electron density given by the sum of atoms in the crystal (the procrystal) (Turner et al., 2015). Electrostatic potentials were mapped onto these surfaces over the range −0.173 to +0.286 au.
Integration was carried out to a resolution of 0.65 Å. Normalisation, including recovery of secondorder harmonic reflections, was carried out using a wavelength range of 0.8-1.7 Å giving a Rmerge of 15.5% for all 21729 reflections. Completeness was 79.3% giving a redundancy of 6.1. The harmonic overlap of reflections results in a maximum possible completeness of 83.3% for Laue diffraction (Cruickshank et al., 1987). Final R1 was 0.0790 (I >2σ(I)) and wR2 was 0.1713 (all data).
The PIXEL calculations show that the interactions between layers individually amount to ca. −4 kJmol −1 . The methyl groups based on C6 and C12 are each positioned in close proximity to three methyl groups in the layers above and below. H…H distances lie between 2.3 and 2.7 Å, but the energy breakdown shows that these are really best considered as whole-molecule dispersion interactions.    Table 3 and Table 4

Room-temperature structure
The structure of L-leucine was modelled using the unit-cell dimensions of (Coll et al., 1986) transformed to the standard setting used by Görbitz  The structure is essentially unchanged from that at 120 K, and again the asymmetric unit consists of two L-leucine molecules as shown in Figure 2(b). Bond lengths are given in Table 2. The hydrogenbonding pattern at room temperature remains similar to that at 120 K. The four methyl groups in the asymmetric unit show a significant disparity in ADPs; the average Ueq values for each group are given in Table 5.

Effects of temperature
Laue diffraction is not capable of determining unit-cell dimensions absolutely using the observed reflection coordinates, only ratios of a:b:c. It is however possible quantify changes in unit-cell lengths on the order of 1%, by observed shifts in the refined instrument spectra produced during the normalisation process (Piltz, 2011) (also see Figure 1). The resulting cell-length multiplier coefficient can be applied to get more accurate unit-cell lengths. A negative shift in the wavelength distribution implies a positive cell-length multiplier coefficient and vice versa.
Calculation of the strain tensor (using the method of Ohashi and Burnham (1973)) using unit-cell The most significant structural change upon cooling to 120 K is the reduction in ADP parameters for the terminal methyl groups, although those of leu-A still remain larger than leu-B. Clearly the enlarged ADPs at 300 K are due at least partially to thermal motion. The significant reduction in this motion is in agreement with the observation from inelastic neutron scattering that significant motion of the CH3 groups occurs above 150 K in the powder form (Façanha Filho et al., 2011).
Upon cooling to 120 K, Ueq(H) values for methyl group based on C6 in leu-B and that based on C5 of leu-A are statistically similar. C6H3 remains enlarged, although only by a small margin above 5 estimated standard deviations.

Conclusions
The structure of the natural amino acid L-leucine has been determined by neutron diffraction for the first time at temperatures of 300 and 120 K. The resulting structures yield geometric parameters with sufficient precision and accuracy for inclusion in restraint libraries of macromolecular structure refinements; the estimated standard deviations on X-H bond lengths range from 0.008-0.03 Å at 300 K and 0.008-0.02 Å at 120 K. Due to the small size or poor quality of L-leucine crystals, the determination of the structure by neutron diffraction has required the application of the modern powerful Laue method.
Calculation of intermolecular interaction energies reveals a pattern of attractive and repulsive interactions. The energies of hydrogen bonds are not correlated with distance but are instead determined by the disposition of positive and negative regions of electrostatic potential. These calculations also reveal a number of important electrostatic interactions, significantly longer than hydrogen-bond distances which are often assumed to be the most important interaction in the analysis of amino-acid structures. As expected for the single-crystal form, no signs of phase transitions were detected during heating or cooling of the single crystal. Cooling leads to minor unit-cell contraction and significantly reduced motion within the alkyl residue as well as minor rearrangements within the hydrogen-bonding network.     (6) -0.007(5) -0.010(4) -0.008 (4)