by Zsolt Bikdia), Rbert Kurdib), Sndor Baloghb), Julianna Szemnc), and Eszter
Hazai*a)
a
) CycloLab Cyclodextrin Research and Development Laboratory Ltd., P.O. Box 435,
H-1525 Budapest
CDs possess for self aggregation, the more likely are they involved in interactions with
carotenoids. These results suggest that self-aggregation of cyclodextrins should be
considered as an important parameter determining complexation in general.
Keywords:
Cyclodextrins
Carotenoids
Aggregation behaviour
Aqueous solubility
Complexation
sequestered inside the cavity of the cyclodextrin. In many cases [13 15] complexation
of drugs by CDs improves their delivery characteristics without interfering with their
activities, because complexation is a rapidly reversible dynamic process. The parent
CDs with appropriate cavity diameter are suitable for complexing many pharmaceutical
compounds, however, they have less than optimal solubility and safety data [15]. Thus,
numerous chemically modified CDs have been developed to improve the solubility
limits of the parent CD [13 15].
The possible factors and the various molecular forces which may play a role in
cyclodextrin complexation has been widely discussed [1], and several hypotheses have
been proposed to account for cyclodextrin complex formations [14][16][17]. Computer
simulations are widely used to rationally explain the experimental findings concerning
inclusion and recognition [18]. Faucci and co-workers have recently developed a
mathematical model for predicting the stability of drug-cyclodextrin complexes with the
aid of molecular modelling techniques [19]. Their model was appropriate for predicting
the stability of CD- small drug complexes, which are able to totally intrude into the
hydrophobic cavity of cyclodextrins. However, in spite of the efforts, the relative
contributions of the different forces involved in the process of complexation are still not
known. Evidence has been shown that interactions among cyclodextrins are also
important factors in determining cyclodextrin complexation behaviour. Recently,
experimental evidence proved the self-aggregation of CDs in water well below the
maximum solubility [20]. Moreover, the self-assembly of CDs has been shown to affect
their aggregation properties.
Previous studies on short-chain analogues of carotenoids, beta ionone retinoids
[21][22] demonstrated the formation of stable inclusion complexes with CDs. It was
shown that the terminal cyclohexene fragment of beta-ionone, which is present in most
carotenoids, has the requisite size for incorporation into the CD cavity. Thus, it seems
likely that carotenoids also form complexes. Recently, experimental evidence of a real
carotenoid-cyclodextrin inclusion complex formation was published [21]. However,
because of the size and shape of carotenoids and CDs, carotenoids will not be totally
buried in the cyclodextrin cavity. Therefore, it is reasonable to assume that even when
forming inclusion complex, a part of the carotenoid molecule is still able to form
interactions with either another carotenoid or cyclodextrin molecule.
In the present work, molecular docking and molecular dynamics calculations
were utilized as a tool for designing carotenoid-cyclodextrin complexes with enhanced
carotenoid water solubility. Typical carotenes and carotenoids were selected for
investigation: open-chain carotene lycopene, -carotene with two -rings and
carotenoids lutein, zeaxanthin and a slightly water soluble carotenoid derivative,
crocetin (Fig. 1). Our experimental results showed that among the parent CDs, only
BCD raises carotenoid solubility at an excess cyclodextrin concentration. Molecular
dynamics calculation was carried out to get deeper insight into the complexation
mechanism. Our calculations revealed that carotenoids are able to form complexes with
CDs with greater than 1:1 stochiometry. Moreover, we showed that the tendency of CDs
to aggregate inversely correlates with their ability to form water soluble complexes with
carotenoids. Finally, our results showed that the CD of choice for complexation with
carotenoids possesses a low affinity for self-aggregation and forms large contact surface
area complex.
Results and Discussion. The complexation of various molecules with CDs is
of high industrial importance as many applications exist in pharmaceutical and
environmental research. Numerous investigations have been carried out on the
mechanism of host-guest inclusion processes in order to get some information about the
forces which lead to the association of a molecule into the cavity of CDs. In our study,
the interactions between the guest molecules and the interior of various CDs were
investigated by both experimental methods and molecular modelling calculations. The
goal of our project was to find the most appropriate CD derivate which enhances the
solubility of the investigated carotenoids. Representative carotenes and carotenoids
selected for investigation were lycopene, -carotene, lutein, zeaxanthin and crocetin
(Fig. 1).
Solubility of carotenoids in parent ACD, BCD and GCD solutions. First, the
complexation behaviour of different carotenoids with parent CDs were examined (Table
1) in order to determine the appropriate cavity size. With the exception of lutein, none
of the investigated carotenoids (lycopene, beta-carotenoid, zeaxanthin, crocetin, Fig. 1)
were able to form soluble complexes with any of the parent CDs. Lutein was the only
carotenoid forming BCD complex that showed a slight solubility, while it was not
soluble in ACD and GCD. These data demonstrate that underivatized CDs are unable to
significantly enhance carotenoid solubility. It was anticipated that molecular modelling
helps predicting which CD is the most likely to form complexes of greater solubility.
Therefore, molecular modelling calculations have been carried out in order to identify
the molecular mechanism underlying the complexation mechanism.
Molecular Docking Studies of Carotenoids to Parent CDs. Molecular modelling
studies were performed in order to determine the relationship between the docking
energies of drug-cyclodextrin complexes and their corresponding solubility. The
docking energies were calculated as the sum of the van der Waals forces and
electrostatic interaction energies. The only carotenoid that showed solubility in any of
the parent CDs was lutein, therefore, this molecule was applied in our docking
calculations. The best energy docking geometries of lutein with ACD, BCD and GCD
can be seen on Fig. 2. In the case of ACD, the size of the lutein ring does not entirely
allow the carotenoid ring to intrude into the cavity, therefore, the ring-OH is not able to
form a hydrogen bridge with a hydroxy group of CDs. Non-specific hydrophobic
interactions are formed between lutein and ACD. In contrast, in case of BCD and GCD,
the lutein ring is buried inside the cavity, and is able to form a hydrogen bond with a
hydroxy group of CD. However, the docking energy of lutein complexes with ACD,
BCD and GCD do not differ substantially (Table 2), indicating that analyzing the
docking energy itself is not sufficient in explaining which CD forms complex with
lutein. Mura and co-workers explained the solubility of inclusion complexes with three
theoretical parameters: intermolecular interaction fields, docking energy and contact
surface [19]. They acquired the most accurate theoretical model for the computational
prediction of the aqueous solubility of the inclusion complex. Therefore, the contact
surface of the complexes as a measure of the degree of fit of the guest molecule into the
host molecule (Table 3), was examined. This was possible first by evaluating the
solvent accessible surface of the single molecules and than by the complexes formed.
After that the decrease in the solvent accessible surface of the complex as compared to
the single molecules was analyzed. Our results showed that the greatest contact surface
is observed in the case of lutein BCD complex. The ring of the lutein cannot totally
intrude into the cavity of ACD. In contrast, there is an unfilled volume in the GCD
cavity, and lutein ring forms interactions with only a part of the CD-ring. Indeed,
solubility experiments showed that lutein dissolves in aqueous BCD solutions, but not
in ACD or GCD (Table 1).
Contact surface calculations of all investigated carotenoid-BCD complex were
carried out. This calculation was based on the assumption that a molecule possessing
similar or higher contact surface than lutein, is able to form water soluble complex with
a BCD derivate. The resulted contact surfaces of the calculated minimum-energy
complexes are shown in Table 4. These data show that zeaxanthin and lutein are the
most promising carotenoids to form complexes with BCDs. Other investigated
carotenoids without a substituent on their six-membered ring and open-chain
carotenoids do not fill up the available space inside the CD cavity.
Stepwise Docking Calculations of Carotenoids to Parent CDs and CD
Derivatives. Many examples are known that the use of derivatized CDs greatly enhance
solubility. Therefore, interaction energies of carotenoids with different BCD derivates
were examined. The docking energies of carotenoids to the different BCD derivates
assuming 1:1 complexes yielded about identical energy values and about identical
contact surface as well (data not shown). This finding is reasonable if we consider that
the substituted groups lie at the surface of the CDs, therefore, are not expected to
influence the interaction with the ligand. Thus, the following factors should be taken
into consideration when explaining the different solubility behaviour of CD derivates as
compared to the parent CD: a) One possibility is that carotenoids form complex with
CDs with greater than 1:1 stochiometry. In this case it is possible that not only the CD
cavity, but also the polar surface of the CD molecule interacts with the carotenoids. b)
The other possibility is that CD self-aggregation influences complexation behaviour.
The CD molecules already in interaction with other CDs are not freely available for
10
interaction with carotenoids. c) Our previous study showed that carotenoids form large
card-pack and head-to-tail carotenoid aggregates [4]. Although this fact is likely to
influence carotenoid-CD complexation in general, it does not explain diffences in
solubility in parent CD and CD derivates for a given carotenoid.
a) The first possibility was examined as follows: another CD molecule was
docked into the lutein- CD 1:1 complex (data not sown). The results showed that the
second CD forms interactions with the first CD molecule rather than with the
carotenoid. Subsequent dockings yielded differently arranged CDs around the
previously docked CDs rather than around the carotenoid molecule. These results
indicate that interaction between the polyene chain of the carotenoid and the polar
surface of the CD are of non-specific nature. Derivatization of BCD does not influence
substantially this non-specific interaction. These results point to the possibility of CDCD complexation influencing solubility rather than more CD molecules interacting with
one carotenoid molecule. b) According to the docking calculations, the effect of CD
self-aggregation should be taken into account in prediction of complexation behaviour
of CD derivates. Indeed, the self-aggregation of CDs in water has recently been
postulated [20].
Molecular dynamics of carotenoid-CD system. Molecular dynamics simulations
were carried out in order to simulate the dynamical behaviour of CDs in the presence of
carotenoids. A simulation box including water, different CDs at a concentration of
11
0.05M, was constructed (in silico CD solution). The simulation was carried out in the
in silico CD solution including a carotene ligand. This way both the self-aggregation
properties of CDs and the possibility of formation of carotene-CD complex with more
than 1:1 stoichiometry could be simultaneously analyzed. Three parallel runs were
carried out for a simulation time of 3 ns. The organization and geometry of BCDs and
carotenoid was analyzed at the end of the run for i) stoichiometry of lutein-BCD
complex ii) BCD aggregation. Every run differed in the stoichiometry of luteinBCD
complex, with 3 12 CD being in interaction distance of lutein. One example of a
complex geometry after 3 ns molecular dynamic run is shown in Fig. 3.
The molecular dynamic simulation revealed that in BCD solution specific
interactions might form between the lutein ring and BCD cavity. Furthermore,
interactions between the polyene chain of lutein and BCDs can be formed, however,
these interactions are non-specific in nature. The exact stoichiometry of the complex
cannot be determined. Besides interacting with the carotenoid, CDs aggregated during
the simulation. The beginning and the end of a molecular dynamics simulation is shown
in Fig. 4. It can clearly be seen that the self-aggregation of CDs influenced the number
of BCDs available for interaction with lutein. The aggregation of CDs in water is also
confirmed by experimental data. Namely, experimental evidence has recently been
reported of BCDs forming aggregates in aqueous solution at low concentrations well
below the maximum solubility [20]. Additionally, it has been shown that aggregation of
12
13
it is reasonable to assume that the number of hydrogen bond donors available in CDs
determine their self-aggregation properties. Indeed, the best energy of interaction was
observed for BCD, followed by HPBCD, DIMEB and RAMEB. The interaction energy
of CDs inversely correlates with their tendency to self-aggregate which in turn inversely
correlates with complexation behaviour. Based on this, the investigated CDs are
predicted to form water soluble complexes with carotenoids in the following order:
RAMEB followed by DIMEB, HPBCD and BCD.
Aqueous Solubility of Carotenoids in Complex with CD Derivatives. Lycopene
and -carotene was not soluble in the aqueous solutions of any of the investigated CDs,
while crocetin and zeaxanthin showed an improved aqueous solubility in RAMEB
(Table 6). Lutein was slightly water-soluble in all investigated BCD derivates. The
solubility rank order of lutein in the investigated BCD derivates was as follows:
RAMEB was the most effective solubilizing agent for lutein, followed by DIMEB,
HPBCD and BCD in accordance with the calculations.
Overall, RAMEB was by far the most efficient BCD derivate that is capable of
forming water-soluble complex with the investigated carotenoids. The calculated
interaction energy of CD dimerization inversely correlated with the results of solubility
experiments (Fig. 6). Thus, our study shows that formation of CD complexes of
carotenoids depends on the aggregation properties of CDs themselves. The less the CDs
aggregate, the more likely are they involved in interactions with carotenoids. The degree
14
of substitution of a BCD derivate correlates with its ability to form complex with
carotenoids.
Conclusion. In this article, we report a study on the complexation behaviour of
carotenoids with CDs using molecular modelling and experimental methods. The aim of
the investigation was to assist designing carotenoids with enhanced water solubility by
different CD derivates. Our results indicate that the investigated carotenoids interact
with BCD derivates but not with ACD or GCDs. Molecular docking studies with
carotenoid cyclodextrin complex with 1:1 stochiometry could explain the capability of
CDs to form complexes with carotenoids as opposed to ACDs and GCDs. However,
docking studies assuming 1:1 carotenoid-CD complex cannot differentiate between the
different complexation behaviour of BCD derivates. Molecular dynamics calculations
revealed that the aggregation properties of CD derivates significantly influence their
complexation behaviour. The less the CDs aggregate, the more likely are they involved
in interactions with carotenoids. Docking calculations identified RAMEB as the BCD
derivate with lowest tendency to aggregate. Indeed, solubility experiments confirmed
that RAMEB complexes show by far the best water solubility. To summarize, our
results showed that complexation of a ligand not buried inside of CD cavity is
dependent on two factors: i) The geometry of the inclusion part of the complex. ii) The
self-aggregation property of the CD itself. Moreover, our results suggest that self-
15
Experimental Part
General. Abbreviations. Cyclodextrin: CD; -cyclodextrin: ACD; cyclodextrin: BCD; -cyclodextrin: GCD; Random methylated -cyclodextrin:
RAMEB; dimethylated -cyclodextrin: DIMEB; hydroxypropylated -cyclodextrin:
HPBCD
Docking Calculations. MMFF94 force field [23] with conjugate gradient method
was used for energy minimization of carotenoids using Spartan04 program. Initial
coordinates of -, - and -CDs were extracted from published x-ray structures of CDprotein complexes (PDB entries: 1CXF, 3CGT, 1P2G, respectively). Gasteiger partial
charges were added to ligand and CD atoms with the aid of Autodock tools [24]. Nonpolar hydrogen atoms were merged, rotatable bonds and solvation parameters were
defined with Autodock tools. Each carotenoid was docked into each CD using the same
parameter set.
Affinity (grid) maps of 1004040 grid points and 0.375 spacing were
generated using the Autogrid program [24]. Preliminary studies showed that without
restrictions Autodock results in an artificial complex, where a cyclodextrin is trapped
16
between the end groups of carotenoids. Box centres were defined to rule out this kind of
complex conformation. AutoDock parameter set- and distance-dependent dielectric
functions were used in the calculation of the van der Waals and the electrostatic terms,
respectively. Docking simulations were performed using the Lamarckian genetic
algorithm (LGA) and the Solis & Wets local search method [25] of Autodock. Initial
position, orientation, and torsions of the ligand molecule were set randomly. All
rotatable torsions were released during docking. Each docking experiment was derived
from 100 different runs that were set to terminate after a maximum of 1,500,000 energy
evaluations. The population size was set to 250. During the search, a translational step
of 0.2 , and quaternion and torsion steps of 5 were applied. CD-CD docking
calculations were performed for BCD, RAMEB, DIMEB and HPBCD. The free energy
of binding was calculated with Autodock [24].
Contact Surface Calculations. The program VEGA ZZ [26] was used to
calculate the contact surface area of carotenoids-cyclodextrin complexes. A probe
radius 1.0 and a density 20 were applied. Surface areas were calculated for CDs and
carotenoids independently and in the docked complex. The contact surface area is
derived from the difference between the sum of independent surfaces and the area of the
complex surface.
Molecular Dynamics Calculations. Molecular dynamics calculations were
carried out using the GROMACS program package [27] to simulate the behaviour of
17
lutein molecule in aqueous solutions of different CDs. The topology files of the solvent
molecules (cyclodextrin) and the solute (lutein) were generated using the Dundee
Prodrg server [28]. Water-CD solvent boxes at 0.05M CD concentration were generated
for CD, BCD and GCD using the editconf and genbox commands of GROMACS. The
solvent boxes were energy minimized by unrestrained MD calculations. The simulation
time was 1 ns. These simulation boxes were applied to solvate lutein and to calculate
the dynamic behaviour of lutein-cyclodextrin systems. Complexes were energy
minimized using the GROMOS force field [29] implemented in the program package.
Unrestrained molecular dynamics simulations (MD) were performed for 3 ns for each
lutein-CD system.
Solubility Experiments. The used cyclodextrins and cyclodextrin derivatives
were the products of CycloLab Ltd. (Budapest, Hungary). The fine chemical grade CD
derivatives are tested, identified and characterized with analytical methods. The average
degree of substitution of statistically substituted derivatives is calculated from NMR
data. For characterization of isomer distribution various methods (TLC, MS, NMR, for
some derivatives, GC, HPLC or capillary electrophoresis) are used. The average
molecular weights of parent and derivatized CDs were determined (ACD: 972.9, BCD:
1135.0, GCD: 1297.2, RAMEB: 1303, DIMEB: 1331.4, HPBCD: 1309.0) The phase
solubility studies were performed in deionised water at 25. Excess amounts (more than
18
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
19
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
20
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
H. J. C. Berendsen, D. van der Spoel, R. van Drunen, Comp. Phys. Comm. 1995,
91, 43.
21
[28]
[29]
22
Table 1. Solubility Data of Carotenoids with a 5% Solution of Parent CDs. The only
complex showed higher aqueous solubility than limit of detection (LOD) is lutein-BCD.
Aqueous solubility (g/ml)
Carotene/CD
ACD
GCD
BCD
Lutein
< LOD
< LOD
0.55 0.05
Lycopene
< LOD
< LOD
< LOD
Beta-carotenoid
< LOD
< LOD
< LOD
Zeaxanthin
< LOD
< LOD
< LOD
Krocetin
< LOD
< LOD
< LOD
23
Table 2. Docked energy of lutein to parent CDs. The calculated difference is negligible.
Cyclodextrin-lutein complex
ACD-lutein
5.43
BCD-lutein
5.41
GCD-lutein
5.44
24
[2]
lutein [2]
Contact
ACD
895.0
1 787.1
1 518.3
268.8
BCD
1 058.5
1 950.6
1 626.9
323.7
GCD
1 177.1
2 069.2
1 793.4
275.8
Lutein
892.1
25
5.32
5.71
5.74
5.58
5.35
1345.0
1626.9
1666.5
1624.5
1597.0
563.3
892.1
915.0
858.7
883.1
1058.5
1058.5
1058.5
1058.5
1058.5
276.8
323.7
307.0
292.7
344.6
26
2.60
HPBCD
+1.81
DIMEB
+3.44
RAMEB
+5.49
27
HPBCD
DIMEB
RAMEB
Lutein
0.87 0.09
10.3 0.6
29.0 1.6
Lycopene
< LOD
< LOD
< LOD
Beta-carotenoid
< LOD
< LOD
< LOD
Zeaxanthin
< LOD
< LOD
20.0 1.2
Krocetin
< LOD
< LOD
3.6 0.3
28
Captions
Fig. 2. Best energy docking results of lutein to a) ACD b) BCD, and c) GCD from two
viewpoints (left and right panels)
Fig. 4. Initial structure and the resulting geometry of cyclodextrin solvent after a 3 ns
MD run. Formation of cyclodextrin aggregates can be observed, which influences their
ability to form complexes with carotenoids.
29