A combined experimental and in silico characterization to highlight additional structural features and properties of a potentially new drug

Article abstract of DOI:10.1016/j.molstruc.2017.06.061

LASSBio-1755 is a new cycloalkyl-N-acylhydrazone parent compound designed for the development of derivatives with antinociceptive and anti-inflammatory activities. Although single crystal X-ray diffraction has been considered as the golden standard in structure determination, we successfully used X-ray powder diffraction data in the structural determination of new synthesized compounds, in order to overcome the bottle-neck due to the difficulties experienced in harvesting good quality single crystals of the compounds. We therefore unequivocally assigned the relative configuration (E) to the imine double bond and a s-cis conformation of the amide function of the N-acylhydrazone compound. These features are confirmed by a computational analysis performed on the basis of molecular dynamics calculations, which are extended not only to the structural characteristics but also to the analysis of the anisotropic atomic displacement parameters, a further information – missed in a typical powder diffraction analysis. The so inferred data were used to perform additional cycles of refinement and eventually generate a new cif file with additional physical information. Furthermore, crystal morphology prediction was performed, which is in agreement with the experimental images acquired by scanning electron microscopy, thus providing useful information on possible alternative paths for better crystallization strategies.

Full text of DOI:10.1016/j.molstruc.2017.06.061

Accepted Manuscript
A combined experimental and in silico characterization to highlight additional
structural features and properties of a potentially new drug
Isadora T.S. Bastos, Fanny N. Costa, Tiago F. Silva, Eliezer J. Barreiro, Lídia M.
Lima, Delson Braz, Giuseppe M. Lombardo, Francesco Punzo, Fabio F. Ferreira,
Regina C. Barroso
PII:
S0022-2860(17)30841-4
10.1016/j.molstruc.2017.06.061
MOLSTR 23947
DOI:
Reference:
To appear in: Journal of Molecular Structure
Received Date: 21 February 2017
Revised Date: 13 June 2017
Accepted Date: 14 June 2017
Please cite this article as: I.T.S. Bastos, F.N. Costa, T.F. Silva, E.J. Barreiro, Lí.M. Lima, D. Braz, G.M.
Lombardo, F. Punzo, F.F. Ferreira, R.C. Barroso, A combined experimental and in silico characterization
to highlight additional structural features and properties of a potentially new drug, Journal of Molecular
Structure (2017), doi: 10.1016/j.molstruc.2017.06.061.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Inference of anisotropic atomic displacement parameters by integrating X-ray powder
diffraction data and a molecular dynamics calculation.

ACCEPTED MANUSCRIPT
A combined experimental and in silico characterization to highlight
additional structural features and properties of a potentially new drug
Isadora T. S. Bastos,a Fanny N. Costa,b Tiago F. Silva,c Eliezer J. Barreiro,c,d Lídia M.
Lima,c,d Delson Braz,e Giuseppe M. Lombardo,f Francesco Punzo,f Fabio F. Ferreirab† and
Regina C. Barrosoa
a. Physics Institute, University of the State of Rio de Janeiro, Rio de Janeiro, Brazil.
b. Center for Natural Sciences and Humanities (CCNH), Federal University of ABC (UFABC), Santo André, SP,
Brazil.
c. LASSBio, Institute of Biomedical Sciences, Federal University of Rio de Janeiro (UFRJ), Rio de Janeiro, RJ,
Brazil.
d. Graduate Program of Chemistry, Institute of Chemistry, Federal University of Rio de Janeiro (UFRJ), Rio de
Janeiro, RJ, Brazil.
e. Laboratory of Nuclear Instrumentation/COPPE, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil.
f. Dipartimento di Scienze del Farmaco, Sezione Chimica,ꢀUniversità degli Studi di Catania, Catania 95125,
Italy.
LASSBio-1755 is a new cycloalkyl-N-acylhydrazone parent compound designed for the
development of derivatives with antinociceptive and anti-inflammatory activities. Although
single crystal X-ray diffraction has been considered as the golden standard in structure
determination, we successfully used X-ray powder diffraction data in the structural
determination of new synthesized compounds, in order to overcome the bottle-neck due to the
difficulties experienced in harvesting good quality single crystals of the compounds. We
therefore unequivocally assigned the relative configuration (E) to the imine double bond and a
s-cis conformation of the amide function of the N-acylhydrazone compound. These features
are confirmed by a computational analysis performed on the basis of molecular dynamics
calculations, which are extended not only to the structural characteristics but also to the
1

ACCEPTED MANUSCRIPT
analysis of the anisotropic atomic displacement parameters, a further information – missed in
a typical powder diffraction analysis. The so inferred data were used to perform additional
cycles of refinement and eventually generate a new cif file with additional physical
information. Furthermore, crystal morphology prediction was performed, which is in
agreement with the experimental images acquired by scanning electron microscopy, thus
providing useful information on possible alternative paths for better crystallization strategies.
Keywords: cycloalkyl-N-acylhydrazone; structure determination; X-ray powder diffraction;
in silico characterization; crystal morphology prediction.
Introduction
The result of drug administration is a complex phenomenon that should be addressed by
both pharmacokinetics and pharmacodynamics. While pharmacokinetics is related to the
physicochemical processes developed by the body to receive a drug, the latter is associated to
the action mechanisms of drugs. In fact, molecules permeate complex biochemical systems
producing intermolecular interactions with biomacromolecules, thus promoting the molecular
recognition of a drug in the biophase.1,2
There are several distinct interactions between a ligand and the target i.e., among others,
hydrophobic interactions, hydrogen bonding, charge transfer complexes, ionic bonds and
covalent bonds.2,3 The spatial arrangement of the structural subunits responsible for these
interactions with the complementary sites of the bioreceptor target includes the
pharmacophoric requisites for its recognition and defines qualitatively the affinity and
selectivity degree of the drug-bioreceptor complex.1
2

ACCEPTED MANUSCRIPT
In order to understand the pharmacological properties as a result of the action mechanism of
a drug, the characterization of the three-dimensional arrangement of the derivatives is
necessary. A better evaluation of the molecular profile in the biophase, as well as the highlight
of the interaction points between drug and receptor can be assessed.1,2 In this context, we
studied the LASSBio-1755 parent compound, which was designed and synthesized as part of
a research program to develop a series of compounds with anti-inflammatory and
antinociceptive activities by structural modifications of the prototype LASSBio-294.4,5
The studied system belongs to a series of new cycloalkyl-N-acylhydrazone compounds
(CANAH). This class of compounds was initially designed by a molecular simplification of
the benzodioxole fragment of LASSBio-294, with subsequent lower homologation; new
molecular changes at the level of the iminic subunit of the NAH function were subsequently
designed, based on the strategy of classic ring bioisosterism, as illustrated in Fig. 1.4-6
Key information related to the structural study of the N-acylhydrazone derivatives is the
knowledge of their conformations and relative configurations since both of these
characteristics are related to their activities.1 N-acylhydrazone compounds have been found
with Z/E geometrical isomers and cis/trans amide conformers.7
3

Fig. 1 – Molecular simplification and design of LASSBio-1755.
ACCEPTED MANUSCRIPT
We performed the present work in order to determine the crystal structure of the LASSBio-
1755 parent compound in the solid state. As we could not harvest this compound as a single
crystal of sufficient size, quality and stability, For this purpose, we used X-ray powder
diffraction (XRPD) data were collected and, by employing a simulated annealing algorithm
implemented into the DASH software,8 we solved the structure, which was then refined by the
Rietveld method.9,10 Information such as intermolecular bonds for the formation of the
crystalline aggregate as well as knowledge on the conformation and configuration of the
studied compound were obtained.6,11-14
The so inferred data were used as a starting point for a molecular dynamics (MD)
calculation, based on a robust approach,15 already successfully proved on several different
systems.16-22 This analysis provided a valuable further tool to study the structural properties of
the compound by means of the atoms three dimensional fluctuations around an equilibrium,
thus allowing – observing their deviation from a perfect sphere – a possible description of the
dominating forces at the atomic level.
Furthermore, the anisotropic atomic displacement parameters (aniso adps) information was
used for an additional Rietveld refinement to finally output a new Crystallographic
Information Framework (CIF) file with more physical information – the adps – embedded.
This is to our knowledge the first attempt, although still in a very basic stage of development,
to implement these data in a XRPD refinement.
This detailed structural picture enabled a crystal morphology prediction, which very well
matched the scanning electron microscopy (SEM) image. These pieces of structural
information, together with a landscape of the dominating intermolecular forces in the unit cell,
4

ACCEPTED MANUSCRIPT
could be particularly useful to try to develop alternative successful strategies to harvest good
quality crystals.
Materials and methods
LASSBio-1755 compound
LASSBio-1755 was obtained in two linear steps using commercial methyl
cyclopropanecarboxylate as the starting material (Fig. 2). The first step consisted in the
hydrazinolysis reaction producing the derivative cyclopropanecarbohydrazide. The second
step was the condensation reaction between cyclopropanecarbohydrazide and
isonicotinaldehyde in good yields (71%). Reactions were routinely monitored by thin-layer
chromatography (TLC) in silica gel (F245 Merck plates) and the products visualized with
ultraviolet lamp (254 and 365 nm).
1H and 13C nuclear magnetic resonance
NMR spectra were recorded on a 200/50 MHz Bruker DPX-200, at room temperature. Peak
positions are given in parts per million (d) from tetramethylsilane as internal standard, and
1
13
coupling constant values (J) are given in Hz. The product showed H and C NMR spectra
according to the assigned structure.
1
NMR H (200 MHz, DMSO, TMS) ppm: 8.60 (d, 2H, J= 4H), 8.16/8.00 (s,1H, N=CH),
7.60 (d, 2H, J=4Hz), 2.67/1.64 (m,1H), 0.99-0.81 (m, 4H).
13
NMR C (50 MHz, DMSO, TMS) ppm: 175.00/169.83 (NC=O), 159.63/142.98 (N=CH),
150.31/150.16 (1C), 142.98/141.65 (1C), 122.81 (2C), 126.89 (2C), 12.94/9.72 (1C),
8.12/7.25 (2C).
5

ACCEPTED MANUSCRIPT
HPLC analysis
The purity determination (99.90%, r.t. 2.23 min) was performed using a Shimadzu –
LC20AD apparatus, with a SHIM-PACK G-CLC-ODS analytical guard column and Kromasil
100-5C18 (4.6 mm x 250 mm) analytical column and the SPD-M20A detector (Diode Array)
at a wavelength of 254 nm for quantification of the analyte in a 1.0 mL min-1 constant flow
rate. The injector was programmed to inject a volume of 20 µL. The solvent systems used
were: 70:30, acetonitrile/water, and isocratic HPLC mode was used.
Fig. 2 – Synthesis steps of LASSBio-1755.
Differential scanning calorimetry (DSC)
The DSC curve was obtained in a differential scanning calorimeter, model DSC Q200, from
TA Instruments, with heat flux using a temperature from 293 K to 523 K and a heating rate of
10 K min-1. A small portion of the sample, with a mass of 3.22 mg, was transferred to an
aluminum crucible, previously weighed. The resulting thermogram was analysed with the TA
Universal Analysis software finding that the sample purity was 99.88%, thus in very good
agreement with the HPLC data.
X-Ray powder diffraction
The LASSBio-1755 sample was gently hand-grinded in a pestle and mortar then passed
through a 10-µm sieve and the XRPD were recorded at room temperature on a STADI-P
powder diffractometer, (Stoe®, Darmstadt, Germany) in transmission geometry by using
6

ACCEPTED MANUSCRIPT
CuKα1 radiation (λ = 1.54056 Å), selected by a curved monochromator Ge (111), with a tube
voltage of 40 kV and a current of 40 mA. The measurement was performed in the angular
range from 6° to 100.485° (2θ) with steps of 0.015° and integration time of 100 s (6°- 28.035°
2θ), 200 s (28.050°- 47.985° 2θ), 400 s (48.000°- 68.985° 2θ) and 800 s (69.000°- 100.485
2θ) at each 1.05°.
Indexing
XRPD data of LASSBio-1755 were used for indexing the first 22 reflections of the pattern
using the Topas-Academic v.5 software program23 and a Pawley fit24 was performed yielding
ꢁ
the following values for a triclinic system with space group ꢀ1: a = 11.372988 Å, b =
9.300565 Å, c = 4.867066 Å, α = 97.46771°, β = 102.75142°, γ = 103.37851° and V =
479.518 Å3. The goodness-of-fit indicator and R-factors25 were: χ2 = 1.898, Rwp = 5.221% and
Rexp = 2.750%.
Structure determination and Rietveld refinement
A global optimization method26,27 implemented into the DASH software program was used
to determine the crystal structure. This step is performed using the information previously
found together with the 3D sketch of LASSBio-1755, created with the program MarvinSketch
version 16.4.25.0, 2016, ChemAxon (http://www.chemaxon.com). Twenty runs (on a total of
2.00 × 108 movements) of the simulated annealing process were globally optimized. In this
process, ten parameters were allowed to vary, with three of them describing the positional
coordinates, four, of which three are independent, describing the molecular orientation and
three flexible torsion angles. In 18 out of 20 solutions the profile-χ2 was less than 5,8
indicating possible success on the structure determination procedure. The best first (profile-
7

χ2/Pawley-χ2: 18.39/3.72 4.9) result was then considered in the Rietveld refinement of the
ACCEPTED MANUSCRIPT
structure.
Following the same procedure as used for the Pawley fit, the background was fit using a 19-
term Chebyshev polynomial and the peak asymmetry was fit by the simple axial divergence
model of Cheary and Coelho.28,29 The peak profiles were modelled by the Double-Voigt
approach30 with anisotropic peak profiles adjusted using a 4-term spherical harmonics.31 Soft
planar restraints on rings and bond distances and angles were restrained to distances and
angles mean values obtained from the MOGUL Cambridge Structure Databank System.32
Isotropic atomic displacements (Biso) were constrained to be equal for all non-hydrogen atoms
and for hydrogen atoms the values were 1.2 times the ones referring to the atoms to which
they are connected. Their fractional coordinates were refined restraining the distances
between H atoms to the ones they are associated to. Initially, their fractional coordinates were
inserted as the ones provided by the Mercury software program33 and then, considering the
mean bond distance values found into the program (C–H = 0.96 Å, N–H = 0.87 Å and O–H =
0.99 Å), the refinement was performed, thus providing the final atomic coordinates of all
atoms.
After the Rietveld refinement, the values found for the reduced34 unit cell parameters were:
a = 4.8594(2) Å, b = 9.3162(5) Å, c = 11.3701(5) Å, α = 73.400(4)°, β = 78.088(3)°, γ =
82.640(4)° and V = 479.70(4) Å3. The goodness-of-fit indicator and R-factors were: χ2 =
2.021, RBragg = 1.330%, Rwp = 5.538% and Rexp = 2.740%. Fig. 3 displays the final Rietveld
plot.
The MD calculations, performed as described below, allowed us to gain two sets of different
isotropic displacement parameters (Uiso) values, corresponding to the existing two
independent formula units per unit cell, after having relaxed the symmetry constraints to get
8

ACCEPTED MANUSCRIPT
an artificial P1 setup (vide infra). They were then arithmetically averaged in order to perform
ꢁ
the final refinement going back to the ꢀ1 space group.
MD simulations
Force field settings
The Materials Studio 7.0 package (Accelrys Software Inc., San Diego, USA) was used to
perform the molecular mechanics (MM) and MD calculations through the Discover program,
using the implemented Amber force field. The charges assigned to each atom were obtained
by the charge equilibration method.35 Settings for non-bond (NB) interactions were atom-
based summation method for the vdW and Ewald summation for Coulomb. The vdW used
cut-off = 15.50, spline width = 5.00 and buffer width = 2.00 Å with a long-range energy
correction (Tail Correction = 15.50 Å), while the Ewald summation parameters were accuracy
= 0.0001 kcal×mol-1 and update width = 5.00 Å. The dielectric value
ε was set to 1.0.
Protocols
The MD simulations started from the energy-minimized structures. The periodic boundary
conditions (PBC) with NPT ensemble (P = 0.0 GPa and T = 298 K) were used for all the runs.
In order to allow the cell to change both shape and volume, we used the Parrinello pressure
control method,36 whereas Berendsen’s thermostat with the default decay constant (0.01 ps)
was used to control the temperature.37 Transients of 2.0 ns were collected, after a 100 ps
equilibration period, with 0.001 ps integration time and a sampling interval of 400 time-steps.
The smart minimizer of the Discover module was used to optimize the structures to a gradient
of 0.1 kcal mol–1.
9

ACCEPTED MANUSCRIPT
Modelling
The structural model submitted to simulation was generated on the basis of the X-ray
structure (unit cell and atomic fractional coordinates). The periodic structural model was built
by considering in the single unit cell the fractional parameters of each atom available in the
ꢁ
CIF. The symmetry related (P1) images were allowed to develop, and then the structure was
converted to P1, by removing any symmetry constraints. This allows to make all the image
atoms in symmetry related positions to move independently, one from another, during the MD
simulation. Finally, the unit cell was developed 4 times along the a-axis, 2 times along the b-
axis and 2 times along the c-axis, to attain a super-cell with a = 19.3113 Å; b = 19.1268 Å; c
= 22.8142 Å;
units.
α = 73.6850°; β = 79.1976°; γ = 83.3145° containing 32 independent molecular
Atomic Displacement Parameters
15
For the detailed and complete description of adps calculation see reference . In order to
estimate the adps anisotropicity we calculated the deviation from a perfect sphere (Ueq) as:
3
2
1
δ
=
λi −Ueq
(
)
∑
Ueq
i=1
(1)
where λi are the eigenvalues of the symmetric matrix Ucif, and Ueq is the arithmetic average of
the three eigenvalues λi. In the case of symmetric ellipsoids (i.e. λ1 = λ2 ≠ λ3), the limit value
of δ for a prolate ellipsoid (λ1 = λ2 << λ3) is 2.44948974 whereas, for an oblate ellipsoid (λ1 =
10

ACCEPTED MANUSCRIPT
λ2 >> λ3) is 1.22474487 (i.e. ½δprolate). This indicates that this parameter allows an immediate
quantification of the ellipsoid shapes and that an ellipsoid with δ > 1.225 is certainly prolate.18
Crystal Morphology prediction
The prediction and the study of the possible crystal morphologies were performed using a
preliminary equilibration protocol, by means of the Forcite module included in the Material
Studio 7.0 package of Accelrys©, adopting the molecular mechanics approximation and the
Compass II Force Field (FF).38 The original experimental CIF file was used as the input for an
energy minimization, performed by means of the Smart Minimizer method. Due to the lack of
significative differences, although the same calculations were also performed on the MD
incorporated file (XRPD+MD.cif, vide infra), these latter results were reported in the
Supplementary Information (SI) only. The morphology protocol is based on the so called
Growth Morphology method. For this purpose, the calculations were performed in fine details
allowing a minimum interplanar distance (dhkl) of 1.000 Å without setting any limit neither to
the values of the three Miller indices nor to the overall number of growing faces.
There are two main contributions to the crystallization energy (Ecr): the energy of each slice
(Eslice), i.e. the energy resulting from the lateral interaction of each formula unit within a slice;
the attachment energy (Eatt) related to the energy released as a consequence of the vertical
interaction of the formula unit with an underlying slice.39 This is summarized as:
Ecr = Eslice + Eatt
(2)
It was demonstrated40 that
G
E
r
att
(3)
11

ACCEPTED MANUSCRIPT
where Gr stands for growth rate. Assuming Ecr to be a constant, an interesting consequence of
equations 2 and 3 is that the bigger Eslice, the smaller Eatt, and the slower their Gr. A
relationship between the molecular interactions – intra- and inter-molecular ones – and the
crystal morphology, can be inferred from the combined analysis of both the Periodic
Boundary Conditions (PBC) present in a slice and the Eatt related to it. Within this framework,
the GM method calculates Eatt as stated in eq. 2. For a detailed description of the method see
references 40-44
.
All these calculations were carried out at 0 K and no surface relaxation takes place.
Furthermore, the surface is considered as a perfect termination of the bulk. The search for
possible solvent accessible voids was performed using the VOID algorithm45 setting a grid of
0.20 Å and a probe radius of 1.20 Å.
Results and discussion
The DSC scan of LASSBio-1755 (reported in Fig S1) shows a well-defined melting
behaviour with a sharp minimum, which easily allows the determination of the corresponding
melting temperature (460.81 K). No further minima or maxima are evidenced by the scan,
thus suggesting, apart from the already reported high purity, a relative good stability shown
by the studied compound. The so inferred enthalpy of fusion ∆Hf is 30.43 kJ mol-1. The back
scan evidenced a recrystallization, thus confirming the stability of the compound up to the end
of the forward temperature ramp (523 K). Apparently, LASSBio-1755 is not showing any
tendency – at least under the experimental setup and conditions we were using – to give rise
to polymorphs.
12

ꢁ
ACCEPTED MANUSCRIPT
The studied compound crystallized in a triclinic crystal system with space group ꢀ1 and its
structure is composed of two formulas units per cell unit (Z = 2), accommodating one
molecule in the asymmetric unit (Z’ = 1). We report two CIFs for this structure: one, named
XRPD.cif (CCDC # 1511968) based on a “standard” experimental approach, i.e. a typical
XRPD refinement after X-ray data collection; a second, resulting as detailed below, by
implementing the MD data (MD.cif) with the ones already present in the XRPD.cif, thus
giving rise to a new CIF with additional physical information such as adps (XRPD+MD.cif –
CCDC # 1555893). Crystal data and details of the structure determination are shown in Table
1.
Table 1 – Details from the Rietveld refinement of the crystal structures of LASSBio-1755. XRPD is the
pure experimental CIF; XRPD+MD the one resulting from the merging and refinement taking into
account the Molecular Dynamics results.
Chemical formula
Formula weight (g mol−1)
Crystal system
Space group
C10H11N3O
189.21
Triclinic
ꢁ
P1 (nr. 2)
XRPD
XRPD+MD
Methodology
4.8594(2);
9.3162(5);
11.3701(5)
4.8659(1);
9.30349(2);
11.3395(2)
a, b, c (Å)
73.400(4);
78.088(3);
82.640(4)
73.348(1);
78.023(2);
82.628(2)
α, β, γ (°)
Volume (Å3)
Z, Z’
479.70(4)
2; 1
479.43(3)
2; 1
ρcalc (g cm−3)
Data collection
Monochromator
2θ range (°)
1.3087(2)
1.3107(2)
Ge(111)
6 - 28.035 / 28.050 - 47.985 / 48.000 - 68.985 /
69.000 - 100.485
0.015 (the detector varied 1.05º at each “time per
step”)
Step size(°)
100 / 200 / 400 / 800
Time per step (s)
13

Refinement
ACCEPTED MANUSCRIPT
6300
Number of data points
999
Number
of
contributing
reflections
134
Number of refined parameters
34
Number of restraints
2.740
5.538
1.330
2.021
2.743
6.669
1.333
2.431
Rexp(%)
Rwp(%)
RBragg(%)
χ2
14

ACCEPTED MANUSCRIPT
Fig. 3 – Rietveld plot of LASSBio-1755 showing the good agreement between the experimental data
(XRPD.cif, black crosses) and the calculated profile (red line). The difference between observed and
calculated data is represented by the blue line. The magenta vertical bars at the bottom indicate the
Bragg peak positions. Magnified region [from 55° to 100° (2θ)] displaying the good agreement
between observed and calculated data.
As specified in the Experimental Section, the MD approach is based on the release of
the whole symmetry for the calculation, in order to avoid any possible restrained atom
oscillation. The agreement between the calculated and experimental unit cell is reported
in Table 2.
Table 2 – Details from the MD generated CIF file of the crystal structure of LASSBio-1755.
MD.cif
XRPD.cif
a, b, c (Å)
α, β, γ (°)
Z, Z’
4.8278;9.5634; 11.4071
4.8594; 9.3162; 11.3701
73.6850;
83.3145
79.1976;
73.400; 78.088; 82.640
2; 1
2; 1
15

ACCEPTED MANUSCRIPT
Fig. 4 depicts the superimposition of the calculated (MD.cif) and experimentally
inferred (XRPD.cif) model. In order to generate the final CIF (XRPD+MD.cif) some
additional cycles of the Rietveld refinement were then carried out taking into account
the fractional coordinates as well as the adps (iso and aniso) generated from the MD
calculations. The latter ones were kept fixed and only the fractional coordinates of the
atoms were allowed to vary, with the hydrogen atoms “ridden” to the atoms they are
connected to. The good visual agreement of the refinement as well as the statistical
parameters (R-factors and goodness-of-fit indicators) allowed us to conclude that this
simple method can be implemented in structure determination from X-ray powder
diffraction data. By comparing the original experimental CIF file – i.e. the
“traditionally” inferred XRPD.cif – with the last output obtained by merging and
refining the X-ray experimental data with the MD one, there are several points that
should be clarified. Apparently, as reported in the (SI), the XRPD.cif seems more
accurate. It also gave rise to minor alerts in the checkcif procedure (see SI). However,
most of these alerts are due to an over-refinement as a consequence of a still non robust
merging step of the MD data with the experimental ones.
What is more, if we carefully consider a procedure – flow chart of the steps to be
added in the refinement – we can obtain even better results. We can therefore conclude
that, although further work should be done, this could be the first step towards a
relevant improvement of the refinement of XRPD data, providing some key details –
aniso adps and the physical information they bear with them – that were always missed
in a typical powder diffraction analysis.
It seems relevant to stress once again that, although we are not facing an improvement
in the figure of merit (FoM) related to the structure refinement, we are still providing a
more realistic CIF file with the information related to the atoms thermal motions which
16

ACCEPTED MANUSCRIPT
can be used, as we will describe by means of our in silico approach, in the second part
of this discussion.46 The almost perfect – and however improved (RMS = 0.0189 of the
latter versus the RMS = 0.0952 of the XRPD.cif and MD.cif comparison) – overlay of
the two structures (XRPD.cif and XRPD+MD.cif) is reported in the SI in Fig. S2. At the
present stage, we are dealing with a final CIF file which is not worse – from the FoM
point of view – than the original experimental one but bears more information.
The crystal structure of LASSBio-1755 enabled us to show structural aspects of the
composite solid phase allowing the characterization of the relative configuration E of
the imine double bond. We also assigned the s-cis conformation of the amide function
of the N-acylhydrazone compound.
The formation of the crystalline aggregate of LASSBio-1755 molecules in solid phase
is shown in Fig. 5 where the hydrogen intermolecular interactions are evidenced by the
cyan dashed lines. The molecules are held together by hydrogen interactions between
the carbonyl and the amidic nitrogen of the N-acylhydrazone function. Atoms involved
in the interactions are: N(1)–H(4)···O(6) (D–H = 0.86(2) Å, H···A = 1.97(2) Å, D···A
= 2.78(2) Å and D–H···A = 156.4(19)º), along –x, 1–y, 1–z, where “D” and “A” stands
for hydrogen donor and acceptor. There is also a weak47 intramolecular H-bond – C(8)–
H(13)···O(6) (D–H = 0.96(3) Å, H···A = 2.42(2) Å, D···A = 2.83(3) Å and D–H···A =
105.8(16)°).
17

ACCEPTED MANUSCRIPT
Fig. 4 – Structures overlay and comparison: XRPD (green) and MD (yellow).
Fig. 5 – The formation of the crystalline aggregate of LASSBio-1755 molecules, with Z = 2,
illustrating the inter- and intramolecular H-bonds (cyan dashed lines).
Our computational approach – in a very good agreement with the formerly reported
DSC data – evidenced a fairly stable unit cell in the considered temperature range.
Alternative successful approaches48-50 were already known to qualitatively and
quantitatively perform an adp analysis.
However, by means of our robust15-22 approach already tested on several different
systems, based on the adps MD simulation, we can monitor the internal molecular
fluctuations, which are directly comparable with those experimentally inferred by
means of diffraction data, thus providing a key tool for validating the simulated
trajectories. Furthermore, there are valuable information inferred by means of this
method: on one hand, we are dealing with powder diffraction data, and these are
18

ACCEPTED MANUSCRIPT
characterized by the presence of isotropic (iso) adp values. The calculated adp are
instead anisotropic (aniso). This aniso characteristic is, on the other hand, the key to
address a detailed structural study in order to analyse how much and in which directions
the atoms vibrate around their equilibrium positions.
In Fig. 6, by using a non-standard magnification of the ellipsoid probability level
(90%), it is possible to emphasize the adps deformation from a perfect sphere – i.e. the
non-realistic, from a physical point of view, iso layout. As evidenced in this figure, the
adp motions in the molecule backbone points mostly in the same direction, while the
two rings show a slightly different movement around their equilibrium.
Fig. 6 – Adp qualitative comparison between computed (Uiso, yellowish) and calculated (Ueq, white, grey,
blue and red for hydrogen, carbon, nitrogen and oxygen atoms respectively).
Fig. 6 also evidences on one hand a fairly good overall estimation of the adp volume
but for the cycloalkylic moiety, i.e. the small ring laying on the left part of the picture,
while on the other hand it shows some difficulties in reproducing the values for
hydrogen atoms – whose handling is actually problematic by means of X-rays.
Without invoking too complicate physical reasoning and justifications – although a
bigger displacement is surely expected for the parts of the molecule which lay further
19

ACCEPTED MANUSCRIPT
from the centre of mass of the molecule itself – a possible reason for the reported adps
magnification is the FF overrating of atoms oscillations. While it could be a possible
quantitative pitfall, it should be borne in mind that the qualitative landscape, and the
consequent adps shape analysis, is instead still valid. In fact, there are many
controversial points of view about the real quantification of the adps magnitude, which
actually depends not only on the numerical methods used for their definition51,52 but
also on the technique used for their measurements.53 However, it is interesting noticing
that, while in the original refinement procedure only Uiso values were used, we are
dealing now with the calculated equivalent isotropic displacement parameters (Ueq).
More in details, we quantified the deviation from a perfect sphere, as detailed in the
Experimental section, and reported this data in Table 3. Interestingly, while the adps of
the backbone atoms are oblate, both the adps of the atoms of the tricycle and of the
heteroaromatic ring are prolate. The presence of prolate ellipsoids on both the side parts
of the molecule should not only be interpreted on the basis of a greater atoms vibrations
while moving far from the centre of mass of the molecule; in fact, a chaotic, random and
stronger oscillation would only imply a bigger adp volume. This is exactly what we can
appreciate in the cycloalkylic side of the molecule. The ellipsoid deformation along a
preferential orientation, which gives rise to the prolate shape, is instead hint of a definite
oscillation along a direction. This is the result of the action of a particular force which
causes the movement of the peripheral parts of the molecule while the backbone atoms
– oblate – keep an almost random movement.
This hypothesis is confirmed by the presence of a complex π-stacking interaction in
the heterocycle which takes place in several slightly different crystallographic
directions, and are favoured by the very short c-axis which allows the presence of
several fragments – and corresponding heterocycles – at an optimal distance (in the
20

ACCEPTED MANUSCRIPT
range 4.590(7)– 4.860(4) Å) to give rise to π-stacking. The adp analysis enabled us to
strengthen the pivotal role performed by the π-stacking which was actually only
deducible by the classic X-ray data interpretation.
Table 3 – Calculated Ueq and their corresponding δ value as defined in eq. (1).
Atom
Ueq (Å2)
δ
0.0362
6
0.0344
0
0.0582
2
N1
N3
N23
O6
C2
C5
C7
C8
C9
C11
C17
C18
C19
C21
H4
H10
H12
H13
H14
H15
H16
H20
H22
H24
0.0506
1
0.0372
0
0.0524
3
0.0357
0
0.0827
0
0.0762
8
0.0339
5
0.0522
4
0.0519
4
0.0658
9
0.0582
5
0.0512
0
0.45018
0.38903
0.66966
0.74555
0.46962
0.79588
0.32347
0.76385
0.72772
0.32744
0.83095
0.65432
0.88289
0.68037
0.81030
1.34139
0.88393
0.70478
0.96972
0.73912
1.10187
1.40984
1.24227
1.33282
0.1021
0
0.0600
4
0.1312
5
21

ACCEPTED MANUSCRIPT
0.1378
0
0.1258
3
0.1180
5
0.0956
3
0.1001
8
0.1201
6
H25
0.10025
1.02864
We then performed a structural analysis at a higher level, i.e. not considering only the
atoms layout in the unit cell, but also the crystal habit. This approach enabled us to draw
some conclusions about the effect on the morphology of all the considered forces that
we formerly analysed in the unit cell. Fig. 7a shows the acquired scanning electron
microscopy (JSM-6010LA, JEOL®, Tokyo, Japan, located at the Multiuser
Experimental Centre, UFABC) image of some LASSBio-1755 crystals, exhibiting a
basic prismatic shape with well-defined and sharp surfaces. In general, the crystals have
thicknesses up to 45 ꢀm and longer lengths.
The very good agreement between the experimentally inferred SEM images and the
computationally derived crystal habit is shown in Fig. 7b. If we would have limited the
morphologically prediction to pure geometrical considerations – such as in the so called
BFDH approach54-56 – we would have only based our analysis on the assumption that
the growth rate of the crystal face is inversely proportional to the interplanar spacing
dhkl. As a consequence, the morphologically important (MI) faces of the crystals would
have been those having the largest dhkl values.54-56 However, this approach completely
neglects the role of the energetics of the systems. Instead, by using the GM approach,
we take into account the attachment energy (Eatt) - the energy released as a consequence
22

ACCEPTED MANUSCRIPT
of the vertical interaction of the formula unit with an underlying slice – thus simulating
the crystal habit as specified in the Experimental section. However, the BFDH data are
reported in the SI (Table S1) for comparison. It could be interesting noticing that not
only the number of MI is less accurate – 7 vs 8 with some relevant faces missing or
wrongly accounted – but there is evidence of a different ranking in terms of
corresponding surface relevance thus allowing us to discard the BFDH approach for a
fine structural analysis.
On the basis of the GM approach we can interpret the total facet area (TFA) values
reported in Table 4 where some higher Eatt values are in fact counterbalanced by the
greater interplanar distance dhkl. This is particularly evident for the last four entries in
Table 4.
Since the calculations were performed in vacuo while the “real” crystals were grown
interacting with solvents, we decided to perform the analysis of the potential role of the
crystallization solvents. This study will give us key information on one hand about the
reasons why some MI faces are overgrown or underestimated by the calculation; on the
other hand it will help us making some practical hypotheses about the solubility of these
crystals and new and potentially more efficient crystallization recipes.16,19
For this purpose, in Fig. 8 we report the results of the computed morphology, drawing
the hkl plane corresponding to the MI crystal faces evidenced in the habit, taking into
account which kind of groups are overlooking the considered face.
23

ACCEPTED MANUSCRIPT
Fig. 7 – Crystal morphology comparison between the SEM collected images (a) and the calculated
crystal habit (b). Miller Indices are reported on the morphologically important (MI) faces.
Table 4 – Morphology predictions for LASSBio-1755 by means of GM calculations. Percentage of
total facet area (TFA) is calculated as 100× (hkl facet area)/(total surface area).
Eatt (Total)
(kcal mol−1)
-16.2911
-15.1595
-16.7300
-35.9330
-37.6315
-36.2963
-32.0516
-33.8921
Total Facet
Area (%)
29.71
29.24
20.17
7.21
hkl
Multiplicity
dhkl (Å)
(001)
(010)
(011)
(101)
(111)
(112)
(102)
2
2
2
2
2
2
2
2
10.6842
8.9036
7.9662
4.6534
4.4862
3.9747
3.9100
3.8375
4.95
0.22
6.27
ꢁ
(111)
2.23
The structure shows a high packing index (70.5%)57 and no residual solvent accessible
voids. We can therefore assume LASSBio-1755 is unable to host solvents without
modifying its crystal structure. Moreover, the attachments of solvent molecules must
take place directly from the outer part of the considered faces. Both this statements are
true whatever crystallization mechanism is considered, the classical nucleation theory or
the two-step mechanism.58,59
Closer inspection of the different MI faces reported in Fig. 7(b), evidences a quite
ꢁ
complex landscape: there are in fact some minor and apical faces – 102 and 111 – with
strong polar characteristic, and few principal faces – 010 and 011 – with a marked
apolar trait. The mixture of ethanol and water – as solvent and non-solvent respectively
24

ACCEPTED MANUSCRIPT
– could have strongly affected the actual growth of the crystals. Even if nearly the 80%
of the TFA is represented by slightly apolar faces – 001, 010 and 011 – the little
discrepancies between the experimentally inferred and computationally derived crystal
habit evidently lies into the small apical faces. The SEM analysis – reported in Fig. 7(a)
– evidences a very simplified picture, which could be achieved, starting from the
computational model, by decreasing the role of 101, 111 and 112. They are, as
ꢁ
compared to the other apical MI faces – 102 and 111 – the less polar ones. Their growth
was particularly inhibited by the use of an excess of an apolar solvent, the ethanol in our
case. This is perfectly compatible with the large abundance of solvent vs. non-solvent
used in our crystallization attempts. By tuning the mixture of polar and apolar
solvent/non solvent, a different crystal habit could be achieved with the consequent
bigger or smaller growth of the corresponding polar or apolar face.
Fig. 8 – MI faces, and their corresponding Miller indices, for LASSBio-1755. For each slab the so called
spacefill layout has been chosen with the atom size range corresponding to the van der Waals radius.
25

ACCEPTED MANUSCRIPT
Conclusions
XRPD is a powerful technique which provides 3-dimensional structure of the
molecule and has been indispensable for characterization of N-acylhydrazone
compounds synthesized in LASSBio that could not be obtained in the form of a good
quality single crystal.
Through XRPD, we unambiguously determined the relative configuration E of the
compound evidencing the structural features of the compound in the solid phase.
Moreover, we observed the s-cis relation between the amide hydrogen and carbonyl
oxygen; it differs from aryl-N-acylhydrazones but it is in very good agreement with
previously reported cycloalkyl-N-acylhydrazones.
On the basis of the X-ray data, we were able to identify possible intermolecular
interactions in the unit cell and thus showing the driving forces responsible for the
formation of the crystalline aggregate of the LASSBio-1755. In this framework, this
analysis was enriched by the adps qualitative and quantitative study, which confirms the
role of the above mentioned intermolecular interactions, adding a different perspective,
on one hand, on the molecule vibrations and stability at the atomic level.
The crystal morphology prediction and it comparison with the SEM data, on the other
hand, eventually suggested at the molecular level, the possible solute-solvent
interactions, which suggest possible alternative route for a successful crystallization.
Furthermore, in this paper we reported what is to our knowledge the first attempt to
implement the in silico inferred aniso adp values in the XRPD data refinement, which
did not provided an improved structure from the numerical point of view – the FoM –
which is only slightly different from the one inferred with the traditional refinement, but
with an additional physical information provided by the adps orientation. One of the
26

ACCEPTED MANUSCRIPT
main apparent pitfall of this procedure seems to be not provided by the method itself,
but by the particular molecular conformation – with an “uncommon” tricycle – which
affects the FF performance, only from a qualitative point of view. Although the
procedure needs to be improved and developed, in order to be better implemented in the
refinement procedure of the XRPD data, there is evidence that the so obtained results
are providing key additional information about the structural features of the studied
compound.
Acknowledgements
We would like to thank the Brazilian agencies from financial support and fellowships:
CAPES (grants no. 88881.062195/2014-01 and 23038.007044/2011-08), INCT-
INOFAR (BR, # 573.564/2008-6), the National Council for Scientific and
Technological Development (CNPq, grants no. 307664/2015-5 and 402289/2013-7).
We also thank the UFABC (BR) for experimental support.
References
1.
Barreiro EJ, Fraga CAM. 2008. Química Medicinal - As Bases Moleculares da Ação dos Fármacos.
2nd ed., Porto Alegre: Artmed. p 243.
2.
3.
Wermuth CG. 2008. The Practice of Medicinal Chemistry. 3rd ed., New York, USA: Academic Press.
Duarte CD, Barreiro EJ, Fraga CAM 2007. Privileged structures: A useful concept for the rational
design of new lead drug candidates. Mini Rev Med Chem 7(11):1108-1119.
4. Silva TF. 2012. Planejamento, síntese e avaliação farmacológica de uma nova série de derivados
cicloalquil-N-acilidrazônicos: análogos de LASSBio-294. Programa de Pós-graduação em Química, ed., Rio de
Janeiro, RJ: LASSBio/UFRJ. p 194.
5.
Sudo RT, Albuquerque EX, Barreiro EJ, Aracava Y, Cintra WM, Melo PA, Noël FG, Sudo GZ, Silva
CLM, Castro NG, Fernandes PD, Fraga CAM, Miranda ALP. 2006. Thienylhydrazon with digitalis-like properties
(positive inotropic effects). ed.: University of Maryland.
6.
Costa FN, Braz D, Ferreira FF, da Silva TF, Barreiro EJ, Lima LM, Colaço MV, Kuplich L, Barroso
RC 2014. Synchrotron X-ray powder diffraction data of LASSBio-1515: A new N-acylhydrazone derivative compound.
Radiation Physics and Chemistry 95:292-295.
7.
isomerization of N-acyl and N-aroylhydrazones. Tetrahedron 42(13):3649-3654.
8. David WIF, Shankland K, van de Streek J, Pidcock E, Motherwell WDS, Cole JC 2006. DASH: a
program for crystal structure determination from powder diffraction data. J Appl Crystallogr 39(6):910-915.
9. Rietveld HM 1967. Line profiles of neutron powder-diffraction peaks for structure refinement. Acta
Crystallogr 22(1):151-152.
Palla G, Predieri G, Domiano P, Vignali C, Turner W 1986. Conformational behaviour and E/Z
10.
Rietveld HM 1969. A Profile Refinement Method for Nuclear and Magnetic Structures. J Appl
Crystallogr 2(2):65-71.
11.
Costa FN, Ferreira FF, da Silva TF, Barreiro EJ, Lima LM, Braz D, and Barroso RC 2013. Structure
Re-determination of LASSBio-294 - A cardioactive compound of the N-acylhydrazone class - Using X-ray powder
diffraction data. Powder Diffraction S491-S509.
27

ACCEPTED MANUSCRIPT
12.
Ferreira FF, Antonio SG, Rosa PC, Paiva-Santos CO 2010. Crystal structure determination of
mebendazole form A using high-resolution synchrotron x-ray powder diffraction data. Journal of Pharmaceutical
Sciences 99(4):1734-1744.
13.
determined using high-resolution synchrotron X-ray powder diffraction. CrystEngComm 13(17):5474.
14. Florence AJ, Shankland N, Shankland K, David WIF, Pidcock E, Xu X, Johnston A, Kennedy AR,
Ferreira FF, Trindade AC, Antonio SG, Paiva-Santos CO 2011. Crystal structure of propylthiouracil
Cox PJ, Evans JSO, Steele G, Cosgrove SD, Frampton CS 2005. Solving molecular crystal structures from laboratory
X-ray powder diffraction data with DASH: the state of the art and challenges. J Appl Crystallogr 38(2):249-259.
15.
Lombardo GM, Portalone G, Colapietro M, Rescifina A, Punzo F 2011. From the x-rays to a reliable
“low cost” computational structure of caffeic acid: DFT, MP2, HF and integrated Molecular Dynamics-X ray
diffraction approach to condensed phases. J Mol Str 994:87-96.
16.
characterization of N-(diethylcarbamothioyl)benzamide triclinic polymorph. J Mol Str 1047:1-8.
17. Lombardo GM, Portalone G, Chiacchio U, Rescifina A, Punzo F 2012. Potassium caffeate/caffeic acid
Lazo Fraga AR, Ferreira FF, Lombardo GM, Punzo F 2013. Experimental and theoretical
co-crystal: the rat race between the catecholic and carboxylic moieties in an atypical co-crystal. Dalton Trans
41:14337-14344.
18.
Lombardo GM, Punzo F 2014. False asymmetry, pseudosymmetry, disorder, polymorphism and
atomic displacement parameters. J Mol Str 1078:158-164.
19.
Lombardo GM, Rescifina A, Chiacchio U, Bacchi A, Punzo F 2014. A top-down approach to crystal
engineering of a racemic D2-isoxazoline. Acta Crystallogr B 70:172-180.
20.
Lombardo GM, Rescifina A, Punzo F 2014. Functional hybrid co-crystals of humic substances: a
growth forecast. CrystEngComm 16:5917-5923.
21.
Lombardo GM, Thompson AL, Ballistreri FP, Pappalardo A, Sfrazzetto GT, Tomaselli GA, Toscano
RM, Punzo F 2012. An integrated X-ray and molecular dynamics study of uranyl-salen structures and properties. Dalton
Trans 41:1951-1960.
22.
Pistarà V, Lombardo GM, Bacchi A, Rescifina A, F. P 2013. Experimental and in silico
characterization of a biological active inosose. Structural Chemistry
23.
Coelho AA, Evans J, Evans I, Kern A, Parsons S 2011. The TOPAS symbolic computation system.
Powder Diffraction 26(S1):S22-S25.
24.
Pawley GS 1981. Unit-cell refinement from powder diffraction scans. J Appl Crystallogr 14(6):357-
361.
25.
Toby BH 2006. R factors in Rietveld analysis: How good is good enough? Powder Diffraction
Aarts E, Korst J. 1991. Simulated Annealing and Boltzmann Machines: a Stochastic Approach to
21(1):67-70.
26.
Combinatorial Optimization and Neural Computing. 2nd ed., Chichester, UK: John Wiley & Sons. p 284.
27. van Laarhoven PJM, Aarts EHL. 1992. Simulated Annealing: Theory and Applications. 4th ed.,
Dordrecht, Holland: Kluwer Academic Publishers. p 198.
28.
Cheary RW, Coelho AA 1998. Axial divergence in a conventional X-ray powder diffractometer. I.
Theoretical foundations. J Appl Crystallogr 31(6):851-861.
29.
Cheary RW, Coelho AA 1998. Axial divergence in a conventional X-ray powder diffractometer. II.
Realization and evaluation in a Fundamental-Parameter profile fitting procedure. J Appl Crystallogr 31(6):862-868.
30.
superconductors. Journal of Research of the National Institute of Standards and Technology 98(3):321-353.
31. Järvinen M 1993. Application of symmetrized harmonics expansion to correction of the preferred
orientation effect. J Appl Crystallogr 26(4):525-531.
Balzar D 1993. X-Ray diffraction line broadening: Modeling and applications to high-Tc
32.
Bruno IJ, Cole JC, Kessler M, Luo J, Motherwell WDS, Purkis LH, Smith BR, Taylor R, Cooper RI,
Harris SE, Orpen AG 2004. Retrieval of Crystallographically-Derived Molecular Geometry Information. J Chem Inf
Comp Sci 44(6):2133-2144.
33.
Macrae CF, Bruno IJ, Chisholm JA, Edgington PR, McCabe P, Pidcock E, Rodriguez-Monge L,
Taylor R, van de Streek J, Wood PA 2008. Mercury CSD 2.0 – new features for the visualization and investigation of
crystal structures. J Appl Crystallogr 41(2):466-470.
34.
Krivý I, Gruber B 1976. A unified algorithm for determining the reduced (Niggli) cell. Acta
Crystallogr A 32:297–298.
35.
Rappe AK, Goddard WA 1991. Charge equilibration for Molecular-Dynamics simulations. J Phys
Chem 95(8):3358-3363.
36.
Parrinello M, Rahman A 1982. Strain fluctuations and elastic-constants. The Journal of Chemical
Physics 76(5):2662-2666.
37.
Berendsen HJC, Postma JPM, van Gunsteren WF, DiNola A, Haak JR 1984. Molecular-Dynamics
with coupling to an external bath. The Journal of Chemical Physics 81:3684-3690.
38.
Sun H 1998. COMPASS: An ab Initio Force-Field Optimized for Condensed-Phase Applications
Overview with Details on Alkane and Benzene Compounds. The Journal of Physical Chemistry B 102(38):7338-7364.
28