Increase in solubility of poorly-ionizable pharmaceuticals by salt formation: A case of agomelatine sulfonates

Article abstract of DOI:10.1021/acs.cgd.7b00805

The search for new solid forms of an active pharmaceutical ingredient (API) is an important step in drug development. Often, an API has a low water solubility, which then leads to low oral bioavailability. For basic or acidic APIs, the rational solution is the preparation of salts. For neutral, poorly ionizable, compounds, the cocrystallization is often the only choice. Agomelatine, a poorly soluble nonionizable amide acting as a melatonergic antidepressant is a typical representative of such class of compounds. Until recently, the only multicomponent forms of agomelatine were cocrystals. In this work, we report the preparation of three salts of agomelatine (hydrogensulfate, mesylate, and besylate) and their solvated forms, along with their crystallographic characterization. Interestingly, the crystal structures of the solvated and nonsolvated hydrogensulfates were determined from the same crystal via a topotactic transformation. In all of the structures, the agomelatine molecule was positively charged with the amide oxygen being protonated. The salt formation was also confirmed by solid state nuclear magnetic resonance measurements and density functional theory calculations. By sulfonate salt formation, up to ～200- times faster dissolution of agomelatine was achieved, which proves that salts might be an attractive alternative even for the poorly ionizable compounds.

Full text of DOI:10.1021/acs.cgd.7b00805

Subscriber access provided by UNIVERSITY OF CONNECTICUT
Article
Increase in Solubility of “Poorly-Ionizable“ Pharmaceuticals
by Salt Formation. A Case of Agomelatine Sulfonates.
Eliška Sko#epová, Daniel Bim, Michal Husak, Ji#í Klimeš, Argyro Chatziadi, Ludek
Ridvan, Tereza Boleslavská, Josef Beránek, Pavel Šebek, and Lubomír Rulíšek
Cryst. Growth Des., Just Accepted Manuscript • DOI: 10.1021/acs.cgd.7b00805 • Publication Date (Web): 29 Aug 2017
Downloaded from http://pubs.acs.org on August 30, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Crystal Growth & Design is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
Increase in Solubility of ”PoorlyꢀIonizable”
Pharmaceuticals by Salt Formation. A Case of
Agomelatine Sulfonates.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Eliška Skořepová a,b, Daniel Bím e, Michal Hušák b, Jiří Klimeš d, Argyro
a,c
Chatziadi , Luděk Ridvan c, Tereza Boleslavská c, Josef Beránek c,
Pavel Šebek c, Lubomír Rulíšeke
a Department of Chemical Engineering, University of Chemistry and Technology Prague, Technicka 3, Prague 6,
Czech Republic
b Department of Solid State Chemistry, University of Chemistry and Technology Prague, Technicka 5, Prague 6,
Czech Republic
c Zentiva k.s., U Kabelovny 130, Prague 10, Czech Republic
d Department of Chemical Physics and Optics, Faculty of Mathematics and Physics, Charles University, Ke Karlovu 3, CZꢀ
12116 Prague 2, Czech Republic
e Institute of Organic Chemistry and Biochemistry, Czech Academy of Sciences, Flemingovo náměstí 2, 166 10 Prague 6,
Czech Republic.
eliska.skorepova@vscht.cz
ACS Paragon Plus Environment

Crystal Growth & Design
Page 2 of 35
1
2
3
ABSTRACT
4
5
6
7
8
The search for new solid forms of an active pharmaceutical ingredient (API) is an important
step in drug development. Often, an API has a low water solubility, which then leads to low
oral bioavailability. For basic or acidic APIs, the rational solution is the preparation of salts.
For neutral, ‘poorlyꢀionizable’, compounds, the coꢀcrystallization is often the only choice.
Agomelatine, a poorly soluble ‘nonꢀionizable’ amide acting as a melatonergic antidepressant
is a typical representative of such class of compounds. Until recently, the only
multicomponent forms of agomelatine were coꢀcrystals. In this work, we report the
preparation of three salts of agomelatine (hydrogensulfate, mesylate and besylate) and their
solvated forms, along with their crystallographic characterization. Interestingly, the crystal
structures of the solvated and nonꢀsolvated hydrogensulfates were determined from the same
crystal via a topotactic transformation. In all of the structures, the agomelatine molecule was
positively charged with the amide oxygen being protonated. The salt formation was also
confirmed by solid state NMR measurements and DFT calculations. By sulfonate salt
formation, up to ~200ꢀtimes faster dissolution of agomelatine was achieved, which proves
that salts might be an attractive alternative even for the poorlyꢀionizable compounds.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
INTRODUCTION
The solid state chemistry is an integral part of the development of new pharmaceuticals, since
different solid forms, such as polymorphs, solvates, salts or coꢀcrystals, have different
physicoꢀchemical properties.1 These may include hygroscopicity, stability or filterability, but
the prime pharmaceutical interest are the dissolution properties.2,3 As the number of low
solubility compounds increases among the newly developed active pharmaceutical ingredients
(APIs) there is an increasing interest to discover new ways to improve their dissolution
properties.
ACS Paragon Plus Environment

Page 3 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
Solubility and dissolution can be increased by various approaches, e.g. by micronization or by
preparation of solid dispersions4, but the method of first choice is the modification of the solid
form by crystal engineering. Salt formation is the obvious route for ionizable compounds, i.e.
those, which contain basic or acidic functional groups.5 For neutral (poorlyꢀionizable)
compounds, coꢀcrystallization represents the alternative approach.6
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In general, owing to their ionic character, salts would be expected to have superior dissolution
and solubility properties in a polar medium. It has been reported that salts can increase the
solubility up to 2000ꢀtimes (delveridine mesylate7), whereas coꢀcrystals usually improve the
solubility only moderately8. For example, in the case of seven various carbamazepine coꢀ
crystals, the solubility was increased 2ꢀ152 times9. However, more recent studies comparing
both salt and coꢀcrystal solubilities proved that such trend is rather not universal. In the case
of clotrimazole10, five salts and two coꢀcrystals were described and the average solubility
increase was 8.68 for salts and 5.0 for coꢀcrystals. For sildenafil11, the average solubility
increase for its five salts was 1.24 and an average of 1.48 for its four coꢀcrystals, respectively.
Finally, eight salts of niclosamide increased solubility by factor of 1.63 in average, while its
only one described coꢀcrystal increased solubility by factor of 2.46.12 Even though the number
of examples given here cannot be considered as exhaustive, it clearly demonstrates that
improvement of solubility properties by salt and coꢀcrystal formation can be in some cases
comparable.
One of the excellent examples of the poorlyꢀionizable compounds is agomelatine (AG,
Figure 1), an active pharmaceutical ingredient used for the treatment of major depressions,
first synthetized by Servier in late early 90’s.13 Agomelatine’s agonistic activity on MT1 and
MT2 melatonin receptors and antagonism at the 5ꢀHT2C serotonin receptors was soon
recognized as the reason of its effectiveness as an antidepressant.14–16 It also exhibits
anxiolytic effects.17 Agomelatine was also studied with respect to treatment of manic
ACS Paragon Plus Environment

Crystal Growth & Design
Page 4 of 35
1
2
3
4
5
6
7
8
9
psychosis, Alzheimer’s disease and generalized anxiety disorder.18 It was approved by the
European Medicines Agency for the treatment of major depressive episodes in 2009. In the
same year, it was approved in Canada and a year later in South America, Australia and other
countries. In contrast, in the United States of America, the drug has not yet been approved for
marketing by the US Food and Drug Administration (FDA). Agomelatine is marketed under
the names of Valdoxan or Thymanax and it is available in the form of filmꢀcoated tablets (25
mg dose).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Up to date, a number of crystal phases of agomelatine have been reported and 18 crystal
structures were published. Four are the structures of the polymorphs of the pure API
(polymorphs I19, II20, III20 and X21; polymorph II is thermodynamically the most stable one21),
11 of them belong to coꢀcrystals (ethyleneglycol20, urea22, glycolic acid22, isonicotinamide22,
methylꢀ4ꢀhydroxybenzoate22, oxalic acid23, pyruvic acid23 and two polymorphs of coꢀcrystal
with hydroquinone23 and with acetic acid20,23), whereas three are the structurally different
iodides24 previously published by our group.
Despite the fact that some amidic salts have already been described, the amide functional
group is still frequently considered as rather nonꢀionizable (or poorly ionizable).25 Only very
few of already reported amidic compounds in solid forms exist as salts (0.3% occurrence in
CSD)24.
Such compounds are mostly found in their neutral state and their ionization is difficult.
Compound is in general considered as ‘nonꢀionizable’ or ‘poorly ionizable’, when it does not
contain functional groups with pKa in the range of 2ꢀ12.26 The pKa of amides is between 20ꢀ25
and the pKa of the conjugate acid is around ꢀ1.27
Until now, only very few pharmaceutically relevant neutral compounds have been
characterized as salts. Two (rare) examples may include carbamazepine28 or dutasteride29.
However, to the best of our knowledge, no data regarding the solubility of the
ACS Paragon Plus Environment

Page 5 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
pharmaceutically relevant amidic salts were published. This concerns also our previous work
reporting on the pharmaceutically not highly relevant iodide salts of agomelatine24, in which
their solubility was not studied.
9
To extend the general knowledge in this area, agomelatine was crystallized with three sulfonic
acids (sulfuric, methanesulfonic and benzenesulfonic). Based on their strength, both coꢀcrystal
and salt formation should be possible in practice. Quite unexpectedly, we observed the
exclusive formation of ionic solids. The five novel crystalline forms were prepared
(AG hydrogensulfate, AG hydrogensulfate methanol solvate hemihydrate, AG mesylate, AG
mesylate monohydrate and AG besylate) and were characterized by singleꢀcrystal Xꢀray
diffraction, powder Xꢀray diffraction, DSC and intrinsic dissolution rate. Last but not least,
our extensive structural and physicoꢀchemical characterization of the compounds is further
complemented by the quantum chemical calculations; the methodology presented here might
be used as efficient predictive tool in design of salts and coꢀcrystals of the APIs.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
EXPERIMENTAL
Materials. Agomelatine (AG, Figure 1), {Nꢀ[2ꢀ(7ꢀmethoxyꢀnaphthalenꢀ1ꢀyl)ethyl]acetamide},
C15H17NO2, is a white solid. The material was provided by Zentiva, k.s., as polymorphs I and
II. The acids used for the preparation of agomelatine salts (Figure 1) and all solvents in p.a.
quality were purchased from SigmaꢀAldrich and were used as received.
ACS Paragon Plus Environment

Crystal Growth & Design
Page 6 of 35
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1 Molecular formula of agomelatine, the acids used and their respective anions
Sample preparation. The hydrogensulfate, mesylate and besylate salts of agomelatine were
prepared by precipitation. Agomelatine was dissolved in ethyl acetate (a nonꢀpolar solvent) at
70 °C. The solution was cooled to r.t. and an equimolar amount of the corresponding acid
(dissolved in a small amount of methanol, when necessary) was added dropꢀwise. Then, the
solution was cooled to 0 °C and the precipitate filtered and dried. In this way, ‘bulk’ powder
samples of the nonꢀsolvated forms of all three salts were prepared.
In order to prepare singleꢀcrystals, the materials were recrystallized in various ways. Highꢀ
quality singleꢀcrystals of only one of the salts (AG besylate) could be prepared without major
difficulties.
AG besylate (AGꢀBS): Agomelatine was dissolved in the excess of ethyl acetate at elevated
temperature and then the equimolar amount of benzenesulfonic acid dissolved in a small
amount of methanol was added. The solution was allowed to cool spontaneously and
crystallize in an open vial. After several days at RT, prismꢀlike singleꢀcrystals were obtained.
AG hydrogensulfate methanol solvate hemihydrate (AGꢀHSꢀI): The preꢀprepared powder AGꢀ
HS was dissolved in methanol at elevated temperature and then the sample was cooled to (ꢀ
ACS Paragon Plus Environment

Page 7 of 35
Crystal Growth & Design
1
2
3
4
5
5) °C. After several days at this temperature in an open vial, needleꢀlike singleꢀcrystals were
obtained.
6
7
8
9
AG hydrogensulfate ansolvate (AGꢀHSꢀII): After the diffraction experiment of AGꢀHSꢀI, the
singleꢀcrystal was desolvated during three days in r.t. The crystal turned white and lost its
transparency, but diffracted reasonably well.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
AG mesylate monohydrate (AGꢀMSꢀI): The preꢀprepared powder AGꢀMS was dissolved in
water – methanol (1:1) mixture at elevated temperature and then the sample was cooled to ꢀ5
°C . After several days at this temperature in an open vial, tileꢀlike singleꢀcrystals were
obtained.
Singleꢀcrystal Xꢀray diffraction. A suitable singleꢀcrystal was glued to a capillary and
mounted on the goniometer. CuKα1,2 diffraction data were collected on Xcalibur PX. Data
collection, cell refinement and data reduction was done by CrysAlisPro CCD30 For structure
solution, SIR9231 and Superflip32 were used. Programs CRYSTALS33, Jana200634 and
Platon35 were used to refine structure and analyze absolute structure. Twin laws were
searched for by ROTAX36. Direct methods were used to solve the structure; the positional and
anisotropic thermal parameters of all nonꢀhydrogen atoms were refined. The hydrogen atoms
were located in a Fourier difference map, but those attached to carbon atoms were
repositioned geometrically. The hydrogen atoms were initially refined with soft restraints on
the bond lengths and angles to regularize their geometry (CꢀH in the range 0.93ꢀ0.98 Å, NꢀH
to value of 0.86 Å, OꢀH to value of 0.82 Å) and Uiso(H) (in the range 1.2ꢀ1.5 times Ueq of
the parent atom). Then, the hydrogen positions were refined as riding. Molecular graphics
were prepared in Mercury37 and Discovery Studio38.
XꢀRay Powder Diffraction. To perform initial analysis, a Philips X’PERT PRO MPD
PANalytical diffractometer with CuKα1,2 radiation (wavelength 1.54180 Å) was used at 40
kV and 30 mA. The samples were scanned at a range 4 ꢀ 40° 2Θ with a step size of 0.017° 2Θ
ACS Paragon Plus Environment

Crystal Growth & Design
Page 8 of 35
1
2
3
4
5
and a step time of 21.32s. The results were analyzed using X’PERT HighScore Plus
software39.
6
7
8
9
To solve the structure of agomelatine mesylate anhydrate (AGꢀMSꢀII), the measurement was
done on PANalytical Empyrean powder diffractometer from 6° to 60° 2Θ with CuKα1,2
radiation. The step size was set to 0.013° 2Θ and the sample was placed in a rotating
capillary. PIXcelꢀ3D area detector was used for the intensity measurement. The structure
solution was done in DASH 3.2 software40. For the solution procedure, the CuKα2
contribution was striped in HighScore Plus software39 , the peak position was determined in
DASH 3.2 and the indexation was done by DICVOL0641. For the structure solution, the
structure of AG mesylate monohydrate (AGꢀMSꢀI) was used as the starting model. The
problem was relatively simple (17 DOF) and, from 100 structure solution runs, 43 runs
converged to the same solution. Rigid body Rietveld refinement of molecular positions and
torsion angles was done in DASH 3.2. Final lattice parameters and peak shape Rietveld
refinement was done by X’PERT HighScore Plus software39. The final Rietveld fit can be
seen in Figure S1 in the Supporting information.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
DFT optimization. The XRPD structure of agomelatine mesylate anhydrate (AGꢀMSꢀII) was
verified by the periodic DFTꢀD geometry optimization using the CASTEP42 program, GGA
PBE functional, TS dispersion correction, kꢀpoint grid 48x48x48, , and the energy cutꢀoff of
550 eV and the convergence criteria on energy of 0.00001 eV, max. force of 0.03 eV/
displacement of 0.001 and max. stress of 0.05 GPa.
Å, max.
Å
All structures were also optimized as isolated (solvated) systems employing Turbomole 6.6
program, BP86 functional along with defꢀTZVP basis set, empirical zeroꢀdamping correction
to dispersion (denoted as +D3), and implicit conductorꢀlike screening model (COSMO) with
ε
= 80.0 for aqueous environment or ε = 6.0 for ethyl acetate. For each compound, neutral, ‘Oꢀ
ACS Paragon Plus Environment

Page 9 of 35
Crystal Growth & Design
1
2
3
4
5
protonated’ and ‘
follows:
N
ꢀprotonated’ AG were investigated. The free energies were evaluated as
6
7
8
ꢄ
ꢎ
ꢀ = ꢁꢂꢃ + ꢁꢅꢆꢇꢈ + ꢉꢊ − ꢉꢊꢋꢌꢍ + ꢀꢏꢐꢃꢑ
,
9
where the Eel is the in vacuo molecular (potential) energy of the system, calculated as a
singleꢀpoint energy (def2ꢀTZVPD basis set) at the PBE+D3/defꢀTZVP equilibrium geometry;
the [EZPVE + RT – RTlnQ] terms were obtained from frequency calculations using the rigidꢀ
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
rotor/harmonicꢀoscillator approximation (for
T = 298.15 K and p = 1 atm) whereas the
solvation energy Gsolv was calculated using conductorꢀlike screening model for real solvents
(COSMOꢀRS) utilizing FINE grid for cavity construction and
BP_TZVPD_FINE_C30_1501.ctd parameter file. The COSMOꢀRS calculations were carried
out on top of BP86/def2ꢀTZVPD/COSMO(
ECOSMO,inf ꢀ Ein vacuo) + ꢁCOSMOꢀRS, where is the COSMOꢀRS chemical potential. All DFT
ε = ∞) single points and Gsolv was determined as
(
ꢁ
calculations (BP86 and PBE) were expedited by expanding the Coulomb integrals in an
auxiliary basis set, the resolutionꢀofꢀidentity (RIꢀJ) approximation.
For the purpose of determination of the protonation site of AG amide bond, the pKa of ‘Oꢀ
protonated’ AG and the pKa of ‘
N
ꢀprotonated’ AG were calculated according to equation:
ꢩ
ꢖ
ꢟ
ꢠꢕꢡꢗꢕ
ꢤꢥꢕ
ꢖꢨ ꢟ
ꢦꢣꢞꢧ
ꢒꢓꢔ = ꢕ ꢗꢕ
,
ꢘꢙꢚꢛꢜꢝꢞ
ꢢꢜꢣꢛ
ꢪꢫ∙ꢃꢬꢖꢭꢮꢟ
where G(AGneutral) and G(AGprot) are the free energies of neutral AG and ‘Oꢀ or Nꢀprotonated’
AG respectively; Gsolv(H+) is the solvation energy of proton in aqueous solution (the value of
ꢀ265.9 kcal/mol was used)43. The atomic coordinates of all optimized structures can be found
in the Supporting information.
The proton transfer pathways in the crystalline phase were obtained using the Vienna Abꢀ
initio Simulation Package (VASP) code, version 5.2.44,45 Standard projectorꢀaugmented wave
potentials were used with planeꢀwave basisꢀset cutꢀoff set to 400~eV. Kꢀpoint sets were
chosen so that the effective supercell side length was at least 18 Å. The PerdewꢀBurkeꢀ
ACS Paragon Plus Environment

Crystal Growth & Design
Page 10 of 35
1
2
3
4
5
6
7
8
9
Ernzerhof exchangeꢀcorrelation functional was used.46 The dispersion corrections of Grimme
and coworkers with BeckeꢀJohson damping were added to improve the accuracy of
description of intermolecular interactions.47,48 The structures were first optimized until all
forces were smaller than 0.015 eV/Å. To make the potential energy scan the hydrogen atom
(proton) was gradually moved in steps towards the carbonyl oxygen. After each step, a
subsequent optimization was performed, in which all the atoms, apart from the oxygen of the
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
OAG group, were allowed to move. Moreover, the hydrogen was only allowed to move in the
coordinates approximately perpendicular to the OAG...Oacid line. The optimized structures,
input files, final energies and scripts used for the calculations) can be found in the Supporting
information.
Solution NMR. Solution NMR was used to determine the chemical purity of the prepared
material and the stoichiometry. Samples were dissolved in d6ꢀDMSO and 1H and/or 13C NMR
spectra were measured by Bruker Avance III™ 500 MHz NMR spectrometer equipped with
Prodigy probe and with repetition delay of 10s.
Solid State NMR. Solid state NMR was used to provide the phase identification and purity of
the prepared materials and also to study the protonation state of agomelatine. 13C and 15N
CP/MAS NMR spectra were measured by Bruker Avance III™ 400 MHz WB (wide bore)
NMR spectrometer equipped with 4 mm probe and with 13 kHz spinning.
Differential Scanning Calorimetry. DSC measurements were performed on the DSC 131
(Setaram, France). The sample was weighed in the aluminum pan (120 µL), covered and
measured in the flow of the argon gas (7.5 ml/min). The investigation was performed in the
temperature range from 20 °C to 250 °C with the heating rate of 10 °C/min. The peak
maximum temperature (Tm) was specified in the DSC result.
ACS Paragon Plus Environment

Page 11 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
Dissolution. Dissolution properties of prepared materials were characterized using intrinsic
dissolution rate (IDR) measurement. IDR describes how fast different solid phases release a
molecule from the crystal lattice into a solution while the influence of particle size is
eliminated. IDR were determined using Sirius inForm (Sirius Analytical, Forest Row, UK).
Disc compacts were prepared by compressing approximately 40 mg of material in a 6 mm
diameter dye under a constant load of 120 kg maintained for 2 minutes. Measurements were
performed using 40 mL of solution at constant pH 2.0 and ionic strength of 160 mM adjusted
by sodium chloride and hydrochloric acid. UV spectra were recorded each 30 seconds.
Absorbance values between wavelength 250 and 350 nm were used to evaluate the amount
released at a given time point. IDR was calculated using zero order linear fit through the
experimental data. Only anhydrous salts were measured. All measurements were performed in
duplicate.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3
Results and Discussion
3.1 Xꢀ Ray Crystallography.
By a crystallization of agomelatine with sulfuric, methanesulfonic and benzenesulfonic acids,
five novel crystalline forms were prepared and their threeꢀdimensional structures were solved.
They included solvated and desolvated AG hydrogensulfate, hydrated and dehydrated AG
mesylate and AG besylate. All structures but AG mesylate anhydrate were solved from
singleꢀcrystal Xꢀray diffraction data. AG mesylate anhydrate was solved from powder data.
Details of the structure solution and refinement can be found in Table 1 and will be discussed
further, separately for each crystalline salt.
Table 1 Crystallographic data
Agomelatine
hydrogensulfate
Agomelatine
hydrogensulfate
Agomelatine
mesylate
Agomelatine
mesylate
Agomelatine
besylate
ACS Paragon Plus Environment

Crystal Growth & Design
Page 12 of 35
1
2
3
4
5
methanol solvate
hemihydrate
AGꢀHSꢀI
monohydrate
form
AGꢀHSꢀII
AGꢀMSꢀI
AGꢀMSꢀII
AGꢀBS
6
7
8
9
C15H18NO2 .
HSO4 . CH4O2 .
0.5H2O
C15H18NO2 .
HSO4
C15H18NO2 .
CH3SO3 . H2O
C15H18NO2 .
CH3SO3
C15H18NO2 .
C6H5SO3
formula
formula wt
color
382.44
341.38
white
357.43
339.41
white
401.48
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
colorless
colorless
colorless
cryst.
morphology
needle
needle
rod
powder
ꢀ
plate
0.13 x 0.18 x
0.81
0.13 x 0.18 x
0.81
0.18 x 0.24 x
0.99
0.10 x 0.26 x
0.50
cryst. size (mm)
temp. (K)
radiation
180
180
200
293
190
Cu Kα
Cu Kα
1.54184
Monoclinic
P 21
Cu Kα
Cu Kα
1.54184
Monoclinic
P 21/c
Cu Kα
wavelength (Å) 1.54184
1.54184
Triclinic
P ī
1.54184
Orthorhombic
P 212121
8.0517(4)
8.4807(3)
28.7902(10)
90
cryst. system
space group
a (Å)
Triclinic
P ī
7.0867(3)
12.9576(5)
20.2964(8)
87.206(3)
87.532(3)
87.370(3)
1858.02(13)
4
8.9755(3)
7.1283(3)
12.4593(5)
90
9.1542(4)
9.4378(4)
11.7447(4)
82.588(3)
67.490(4)
72.856(4)
895.58(7)
2
12.2894(3)
15.1147(5)
9.0056(2)
90
b (Å)
c (Å)
α (deg)
β (deg)
93.713(4)
90
97.454(1)
90
90
γ(deg)
90
volume (Å3)
Z
795.48(5)
2
1658.66
4
1965.91(14)
4
density (g/mL)
no. unique
reflns
1.367
1.425
1.325
1.359
1.356
12672
4868
15063
ꢀ
5999
no. obs. reflns
no. of params
6493
515
2495
217
3697
229
ꢀ
3410
261
ꢀ
Rp = 0.0538
Rwp = 0.0715
ꢀ
R_all, wR2_all
0.0509, 0.1098
0.0422, 0.1014
ꢀ0.41, 0.39
0.0462, 0.1113
0.0420, 0.1026
ꢀ0.35, 0.22
0.0673, 0.1526
0.0566, 0.1360
ꢀ0.44, 0.44
0.0438, 0.1103
0.0411, 0.1047
ꢀ0.34, 0.24
R_obs, wR2_obs
ꢁρmin, ꢁρmax (e
Å−3)
ꢀ
GOF
0.9795
0.8845
0.9323
ꢀ
0.9752
CCDC number
1545354
1545355
1545356
1545357
1545353
Agomelatine Hydrogensulfates.
Crystallizing AG with sulfuric acid in ethyl acetate, AG hydrogensulfate (AGꢀHS) is readily
formed. However, the singleꢀcrystal growth turned out to be particularly challenging since the
only highꢀquality singleꢀcrystal did not correspond to the starting AGꢀHS (Figure 3 ꢀ d), but
the MeOH solvate hemihydrate (AGꢀHSꢀI) was formed instead. After a subsequent
desolvation ꢀ without a significant loss of singleꢀcrystal quality ꢀ the second diffraction
ACS Paragon Plus Environment

Page 13 of 35
Crystal Growth & Design
1
2
3
4
5
experiment was carried out. The new structure corresponded to the desired AG
hydrogensulfate ansolvate (AGꢀHSꢀII, Figure 2 and Figure 3 ꢀ h).
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 2 Overview of the formation of hydrogensulfates of AG
The molecular structures of the two AG hydrogensulfates (asymmetric units, Hꢀbonding,
molecular packing and a theoretical XRPD compared to experimental XRPD of the starting
AGꢀHS, labeled as AGꢀHSꢀII exp., because it is the form it corresponds to) are depicted in
Figure 3. AGꢀHSꢀI crystallized in the triclinic system with the space group
P ī. In its
asymmetric unit, it has two molecules of protonated agomelatine, two hydrogensulfate anions,
two molecules of methanol and a single water molecule (Figure 3 ꢀ a), making it a methanol
solvate hemihydrate.
It can be seen that hydrogen bonding system is quite complicated (Figure 3b). However, a
basic structural motif is characterized by the two hydrogensulfate homodimeric units in the
center that are bridging two AG molecules. There exist two basic variants of this motif
depending on the presence or absence of water. Each motif either contains two (right in
Figure 3 ꢀ b) or none (left in Figure 3 ꢀ b) molecules of water. From the structural data, it can
be inferred that the Hꢀbonding with the water molecules is stabilizing factor, since their
absence leads to the disorder in the hydrogensulfate positions.
ACS Paragon Plus Environment

Crystal Growth & Design
Page 14 of 35
1
2
3
AGꢀHSꢀII crystallized in the monoclinic system with the space group
P 21. In the asymmetric
4
5
6
7
8
unit, it contains a molecule of protonated agomelatine and a hydrogensulfate anion (Figure 3 ꢀ
e). Hydrogen bonding pattern (Figure 3 ꢀ f) is quite similar as in AGꢀHSꢀI, but the
agomelatine and hydrogensulfate molecules interact directly due to the lack of any solvent.
Molecular packing is very similar in both structures (Figure 3 – c, g). In both, we can observe
hydrophobic and hydrophilic layers. Interestingly, the symmetry change from triclinic to
monoclinic (Table 1) upon the desolvation implies that topotactic rather than singleꢀcrystal to
singleꢀcrystal transformation takes place.49 Such transition proceeds via nucleation and
growth of a new phase (ansolvate), with the crystallites (domains) that are more or less
perfectly oriented, simulating the diffraction pattern of a single crystal. This is further
evidenced by the lower similarity of the XRPD patterns of the two forms (Figure 3 – d) than
would be expected for SCSC.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The traces of the original solvate centrosymmetric structure could be found in the crystal of
the ansolvate, because, though having an enantiomorphic space group, it had to be refined as a
centrosymmetric racemic twin with the twin matrix of (1 ꢀ1 1) and the refined twin ratio of
0.59 : 0.41.
ACS Paragon Plus Environment

Page 15 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3 Comparison between agomelatine hydrogensulfates. A MeOH solvate hemihydrate
AGꢀHSꢀI (aꢀd) and an ansolvate AGꢀHSꢀII (eꢀh). Asymmetric unit in a, e; Hꢀbonding in b, f;
crystal packing in c, g; agreement in the theoretical and experimental XRPD in d, h.
ACS Paragon Plus Environment

Crystal Growth & Design
Page 16 of 35
1
2
3
Agomelatine Mesylates
4
5
6
7
8
Many aspects of the crystallization of the AG mesylates are similar to AG hydrogensulfates.
By crystallization of AG with methansulfonic acid, agomelatine mesylate (AGꢀMS) is
prepared. In analogy with the AGꢀHS, a singleꢀcrystal structure of a solvated form was
observed (a monohydrate, AGꢀMSꢀI), which did not correspond to the bulk material (Figure 4
– d shows the lack of similarity in the XRPD patterns). Unlike in the case of AGꢀHS, we were
not able to obtain a singleꢀcrystal of the anhydrous form (AGꢀMSꢀII), neither by
recrystallization, nor by SCSC transformation. The crystal structure of the AGꢀMSꢀII was
determined from the powder Xꢀray diffraction data. Figure 4h shows the final Rietveld plot
and the difference curve.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
To verify the structural determination of agomelatine mesylate anhydrate (AGꢀMSꢀII)
obtained employing powder Xꢀray diffraction technique, the DFTꢀD geometry optimization
was carried out. The optimization was done in a 2ꢀstep way as suggested by J. Streek and M.
A. Neumann50. First, the geometry of the structure was optimized, followed by the cell
parameters optimization. To be able to compare original geometry and DFTꢀD optimized
geometry, Cartesian rootꢀmeanꢀsquare displacement (RMSCD) as defined in ref.50 was
calculated in Crystal CMP software51. The RMSCD values of 0.0524 Å (cell fixed, Hꢀatom
excluded) and 0.0753 Å (cell optimized, Hꢀatoms excluded) are both less than 0.25 Å, which
is typical for correctly solved structures50. Therefore, the structure solution of agomelatine
mesylate anhydrate from the XRPD data can be considered as reliable. It should be mentioned
that the positions of H atoms obtained by DFT confirmed the salt characteristic of the solid
form.
The AGꢀMSꢀI crystallized in the triclinic system with the space group
P ī. In its asymmetric
unit, it has a molecule of protonated agomelatine, a mesylate anion and a single water
ACS Paragon Plus Environment

Page 17 of 35
Crystal Growth & Design
1
2
3
4
5
molecule (Figure 4a), making it a monohydrate. Hydrogen bonding in AGꢀMSꢀI (Figure 4b)
can be described as an infinite ladderꢀlike chain.
6
7
8
The AGꢀMSꢀII was found to crystallize in the monoclinic system with the space group
P 21/c.
9
In the asymmetric unit, it contains a molecule of protonated agomelatine and a mesylate anion
(Figure 4e). Hydrogen bonding in AGꢀMSꢀII (Figure 4f) forms a simple infinite chain with
alternating agomelatine and mesylate molecules. As can be seen in Figures 4c, g the
hydrophobic and hydrophilic layers can be identified in both AGꢀMSꢀI and AGꢀMSꢀII. At the
same time, the two structures exhibit a completely different molecular packing, which
precludes even the possibility of the SCSC transformation.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment

Crystal Growth & Design
Page 18 of 35
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4 Comparison of agomelatine mesylates. A monohydrate AGꢀMSꢀI (aꢀd) and an
anhydrate AGꢀMSꢀII (eꢀh). Asymmetric unit in a, e; Hꢀbonding in b, f; crystal packing in c, g;
agreement in the theoretical and experimental XRPD in d, h.
ACS Paragon Plus Environment

Page 19 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
Agomelatine Besylate
AG crystallizes with benzenesulfonic acid (AGꢀBS) in the orthorhombic system with the
space group 212121. In the asymmetric unit (Figure 5a), the structure contains a protonated
P
9
agomelatine molecule and a besylate anion. Hydrogen bonding involves amidic group of
agomelatine and the sulfate group of the besylate anion. They form infinite chains (Figure 5b)
and result in hydrophobic and hydrophilic layers in the molecular packing (Figure 5c). The
solved crystal structure of AGꢀBS corresponded to the powder bulk material (Figure 5d). No
hydrated or solvated phases of AG besylate were found.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment

Crystal Growth & Design
Page 20 of 35
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5 Structural features of agomelatine besylate. Asymmetric unit in a; Hꢀbonding in b;
crystal packing in c; agreement in the theoretical and experimental XRPD in d.
ACS Paragon Plus Environment

Page 21 of 35
Crystal Growth & Design
1
2
3
Comparison of Agomelatine Crystal Structures
4
5
6
7
8
9
The five new crystal structures of agomelatine were compared with all of the 18 various
previously published structures, mostly concentrating on its molecular conformations and
crystal packing. A detailed comparison can be found in the Supporting information.
Agomelatine was found to exist in three general conformations related to the torsion angle of
the side chain (see Figures S2 and S3 in the Supporting Information).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The similarities and differences in the molecular crystal packing were analyzed employing a
packing similarity tree diagram. Figure S4 in the Supporting Information shows the diagram
for the AG structures calculated by the CrystalCMP software51.
The structures were generally very dissimilar and mostly, do not form structural types. This
lack of similarity is quite unusual, because, when a larger number of crystal structures of the
same compound is studied, typically, they are comprised of several ‘isostructural’ families3,51–
53. Despite the general lack of similarity, we have found that one of the here presented
structures (AGꢀHSꢀI) is surprising similar to agomelatine hydriodide trihydrate (Figure 6).
ACS Paragon Plus Environment

Crystal Growth & Design
Page 22 of 35
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 6 Similar crystal packing of agomelatine molecules in AGꢀHSꢀI (b, violet in Figure a)
and AG hydriodide trihydrate (c, red in Figure a). Molecular overlay (a) calculated in
CrystalCMP.
We find it quite surprising that two structures with such a different composition can have a so
similar packing. The sterical end electronic properties of the guest molecules are probably just
right to promote the same packing of the agomelatine molecules.
3.2 Ionization of Agomelatine
Although the amidic group can be protonated in the presence of a strong acid, this is quite rare
in the solid state and thus only 0.3 % of amides found in the CSD are in a protonated state.
Such product of protonation would be expected to be rather unstable (unisolable) or prone to
hydrolysis.54
ACS Paragon Plus Environment

Page 23 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
In the present study, we have prepared solid forms of AG with various sulfonic acids. In all
cases, we have surprisingly obtained the protonated forms of AG, in contrast with the
expected coꢀcrystals (coꢀcrystal formation was expected especially with benzensulfonic acid).
Therefore, we further complement the data from Xꢀray crystallography (previous section)
with other experimental and theoretical methods in order to provide an additional evidence to
support the formation of protonated solid forms. These data are also invaluable in the better
understanding of the process of protonation of poorlyꢀionizable compounds in general.
One of the simplest indications, whether the multiꢀcomponent solid form is a salt or a coꢀ
crystal, is the ꢁpKa rule. It was shown that if the difference between the pKa of the base and of
the acid is greater than 4, the material is most likely a salt. If the ꢁpKa is smaller than ꢀ1, the
material is most likely a coꢀcrystal55. Between these two limits, the formation of both salts and
coꢀcrystals can be expected. The probability of a compound being salt or coꢀcrystal is linear
and can be calculated as55
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ꢖ
ꢟ
ꢯꢐꢰꢏ ꢱꢲꢱꢳ. , % = −17 ∆ꢒꢓꢔ + 72
ꢖ
ꢟ
ꢯꢐꢰꢏ ꢴꢵꢋꢶ, % = 17 ∆ꢒꢓꢔ + 28
In the case of AG sulfonates reported in this work, the ꢁpKa values are in between the limits
for all three of them (Table 2) and both forms are therefore possible. As can be seen in Table 2,
salt formation might be expected for AG hydrogensulfate, whereas coꢀcrystal formation is
more likely to occur for AG besylate. For AG mesylate both forms might have the same
probability.
Table 2 The estimated values of ꢁpKa of AG sulfonates and the related probabilities of the
forms being either salt or coꢀcrystal.
Coꢀcrystal
probability
26 %
Salt probability
*
ꢁpKa
Form
2.7
1.2
0.1
74 %
48 %
30 %
AG hydrogensulfate
AG mesylate
52 %
70 %
AG besylate
ACS Paragon Plus Environment

Crystal Growth & Design
Page 24 of 35
1
2
3
* All the pKa values were calculated by ACD Labs as available in SciFinder.
4
5
Although, the simple ꢁpKa rule may be indicative in many cases, more advanced methods are
6
7
8
9
usually needed to elucidate the true ionization state of the solid forms. We employed the
combination of three methods: analysis of the crystal structures, DFT calculations and solid
state NMR.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The Xꢀray crystallography, provided two independent proofs that agomelatine is protonated on
the amidic oxygen (Figure 7a, b). First, utilizing our highꢀquality data, the position of the
acidic hydrogen is evident from the calculated electron density. For agomelatine
hydrogensulfate (AGꢀHSꢀII, Figure 7a), we can directly observe that the proton is located on
the amidic oxygen of agomelatine. Second, the experimentally observed changes in the CꢀO
and CꢀN bonds clearly indicated the decrease of the bond order in the former and increase of
the bond order in the latter, which is perfectly consistent with the protonation of the amidic
oxygen (Figure 7b, Table 3).
ACS Paragon Plus Environment

Page 25 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 7 Protonation of agomelatine in AG hydrogensulfate (AGꢀHSꢀII) compared with pure
AG (polymorph II) as evidenced by SXRD (a, b) and ssNMR (c, d): a) electron density; b)
change in amidic bond lengths; c) 15N ssNMR; d) 13C ssNMR. Fourꢀ and fiveꢀpointed stars
indicate N and C atoms in the amidic group, respectively. Green stars indicate data/atoms for
pure AG, pink for AGꢀHSꢀII.
ACS Paragon Plus Environment

Crystal Growth & Design
Page 26 of 35
1
2
3
Table 3 Amidic group bond lengths as found in CSD, extracted from crystal structures of
4
5
6
agomelatine and the calculated results.
7
8
9
Form
CꢀN [Å]
1.312
C=O [Å] (CꢀN)ꢀ(CꢀO) [Å]
AG ꢀ pure polymorph I
AG ꢀ pure polymorph II
1.223
1.247
0.089
0.087
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1.334
CSD: neutral amides (median,
96,162 structures)
1.354
1.226
0.128
Calc.: neutral AG
AG hydrogensulfatea
AG mesylateb
1.36
1.24
0.12
1.290
1.302
1.291
1.289
1.284
1.290
0.001
0.018
0.001
AG besylate
CSD: amide salts (median,
244 Oꢀprotonated structures)
1.309
1.298
0.011
Calc.: Oꢀprotonated AG
1.31
1.58
1.32
1.19
0.01
0.39
Calc.: Nꢀprotonated AG
a averages of AGꢀHSꢀI and AGꢀHSꢀII; b data corresponding only to AGꢀMSꢀI
Further independent evidence was provided by the quantum chemical calculations. The
calculated relative pKa values differ by approximately 8 units: ꢀ0.7 for the ‘
structure vs. ꢀ8.5 for the ‘ ꢀprotonated’ structure (c.f. Figure S5).
For direct comparison with the experimental bond lengths, the neutral, the ‘
the ‘ ꢀprotonated’ AG structures were optimized at the BP86/defꢀTZVP level employing
Oꢀprotonated’
N
O
ꢀprotonated’ and
N
implicit solvation model COSMO (with ε = 6.0 corresponding to the dielectric constant of
ethyl acetate). Calculated bond distances are summarized in the Error! Reference source not
found. above. The calculations clearly show that the protonation of agomelatine leads to the
change in the bond orders. Expectedly, the protonation of the oxygen atom leads to a decrease
of the bond order of CꢀO bond (large CꢀO bond distance), whereas the character of the CꢀN
bond is closer to the double bond. The opposite holds true, when agomelatine is protonated on
nitrogen.
The ionization state of agomelatine was also confirmed based on the calculations in periodic
(solid state) quantum chemical calculations at the DFT (PBE+D3) level of theory employing
the VASP package, v. 5.2 (c.f. Computational Details above). The data are summarized in
ACS Paragon Plus Environment

Page 27 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
9
Figure 7. All geometry optimizations for AGꢀBS, AGꢀMS and AGꢀHS lead to the exclusive
formation of protonated agomelatine. Also, moving on the potential energy surface from salt
to coꢀcrystal (i.e. utilizing scan procedure along the OAG–H…Oacid with keeping OAG–H bond
fixed in each step and with relaxation performed for the rest of the system) we did not observe
any other local minimum than the one corresponding to the protonated agomelatine. In fact,
the relative energies of coꢀcrystals are several kcal.molꢀ1 higher than the corresponding
minima. The biggest difference in the energy of coꢀcrystal and salt is observed for AGꢀHS
system, which is also in accordance with ꢁpKa rule. The relative energies of AGꢀMS and AGꢀ
BS coꢀcrystals are comparable and approximately 5 kcal.molꢀ1 higher than the respective
agomelatine salt.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 8 The relative energy of the simulated salt coꢀcrystal continuum for the agomelatine
hydrogensulfate (AGꢀHS), besylate (AGꢀBS) and mesylate (AGꢀMS).
Finally, the solid state NMR (ssNMR) was employed to provide further experimental evidence
concerning the ionization state of the studied molecules. In Figures 7c, d, an example of the
ACS Paragon Plus Environment

Crystal Growth & Design
Page 28 of 35
1
2
3
4
5
15N and 13C spectra of AGꢀHSꢀII and of pure agomelatine are depicted, whereas all ssNMR
data are summarized in Table 4.
6
7
8
9
Table 4 The chemical shift of amidic C and N in the corresponding ssNMR and compared to
pure AG (the average between AG polymorphs I and II).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
chemical
shift ꢀ15N
[ppm]
ꢁ 15N (vs.
pure AG)
[ppm]
ꢀ21.61
ꢀ27.81
ꢀ20.11
ꢀ
chemical
shift ꢀ13C
[ppm]
ꢁ 13C (vs.
pure AG)
[ppm]
ꢀ3.95
ꢀ5.35
ꢀ7.45
ꢀ
Solid form
AG hydrogensulfate (AGꢀHSꢀII)
AG mesylate (AGꢀMSꢀII)
AG besylate (AGꢀBS)
ꢀ259.50
ꢀ253.30
ꢀ261.00
ꢀ282.41
ꢀ279.80
174.60
176.00
178.10
171.60
169.70
AG ꢀ pure polymorph I
AG ꢀ pure polymorph II
ꢀ
ꢀ
As can be seen from both Figure 7 and Table 4, the transformation between the two forms ꢀ the
pure agomelatine forms and their sulfonates ꢀ is accompanied by the significant
increase/decrease of the chemical shift. Interestingly, the change of the chemical shift of the
amidic C and N does not seem to correlate either with the ꢁpKa value or with the difference in
the bond lengths in the crystal structures. Moreover, the change of the chemical shift is
different and uncorrelated for the 15N and 13C atoms.
In summary, all forms can be clearly described as salts, even though the coꢀcrystal formation
was expected for AG besylate (or more precisely, predicted by the ꢁpKa rule). This
observation is justified by the crystal structures and packing of AG solid forms, which is more
energetically favorable for the molecules in an ionic forms.
3.3 Dissolution
The need to increase solubility and dissolution is often the driving force to search for salts and
coꢀcrystals, as they both often offer superior properties compared to the parent compound.
Therefore, here we provide the experimental data about intrinsic dissolution rate (IDR) of the
agomelatine sulfonates compared to the thermodynamically most stable pure polymorph of
ACS Paragon Plus Environment

Page 29 of 35
Crystal Growth & Design
1
2
3
4
5
6
7
8
AG, form II. IDR is used for a direct comparison between the speeds of dissolution of
different solid phases, without the influence of particle size distribution. It is expressed as the
mass dissolved per unit time and area.
9
The IDR results for AG sulfonates compared to AG form II can be found in Figure 9 and
Table S1 and Figure S6 in the Supporting information. We can see that all three salts provide
an increase in dissolution. Agomelatine hydrogensulfate dissolved 3.2 times faster than AGII;
agomelatine besylate 51.4 times faster and agomelatine mesylate an astonishing 200 times
faster.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 9 Intrinsic dissolution of agomelatine sulfonates compared to AG polymorph II
We have to note, that the extremely fast dissolution of the last two mentioned samples also
effected the precision of the measurement itself. Only several points at the beginning of the
dissolution experiments can be considered, because the surface of the sample pellet quickly
became uneven. Therefore, the presented values should be taken as tentative. Nevertheless, it
is clear that the salt formation increased the dissolution of agomelatine in a significant way.
ACS Paragon Plus Environment