Stereolability of dihydroartemisinin, an antimalarial drug: A comprehensive kinetic investigation. Part 2

Article abstract of DOI:10.1021/jo102392p

Artemisinin or qinghaosu has now largely given way to the more potent dihydroartemisinin (DHA, 1) and its derivatives in the treatment of drug-resistant malaria, in combination with other classical antimalarial drugs. DHA is obtained by NaBH₄ reduction of artemisinin and contains a stereochemically labile center at C-10, which provided two lactol hemiacetal interconverting epimers, namely 1α and 1β. In the solid state, the drug consists exclusively of the β-epimer; however, upon dissolution, the two epimers equilibrate, reaching different solvent-dependent ratios with different rates. Such equilibration also occurs in vivo, irrespective of the isomeric purity at which the drug would have been administered. The aim of this study was then to achieve an in-depth understanding of the kinetic features of the α/β equilibration. To this purpose, free energy activation barriers (ΔG^a) of the interconversion were determined as a function of both general and specific acid and base catalysts, ionic strength, and temperature in different solvents by dynamic HPLC (DHPLC). In hydro-organic media, the dependence of ΔG^a on temperature led to the evaluation of the related enthalpic and entropic contributions. Theoretical calculations suggested that the rate-determining step of the interconversion is not the ring-opening of the cyclic hemiacetal but the previous reversible deprotonation of the individual epimers (base-catalyzed mechanism). The whole findings may contribute to shed some light on the mechanism of action and/or bioavailability of the drug at the molecular level.

Full text of DOI:10.1021/jo102392p

ARTICLE
pubs.acs.org/joc
Stereolability of Dihydroartemisinin, an Antimalarial Drug:
A Comprehensive Kinetic Investigation. Part 2
†
‡
‡
‡
†
Walter Cabri, Ilaria D’Acquarica, Patrizia Simone, Marta Di Iorio, Michela Di Mattia,
,
‡
†
§
,‡
†
Francesco Gasparrini,* Fabrizio Giorgi, Andrea Mazzanti, Marco Pierini,* Marco Quaglia, and
Claudio Villani
‡
†
Analytical Development, R&D Department, Sigma-Tau S.p.A., Via Pontina km 30.400, 00040 Pomezia, Italy
‡
Dipartimento di Chimica e Tecnologie del Farmaco, Sapienza Universit ꢀa di Roma, P. le Aldo Moro 5, 00185 Roma, Italy
§
Dipartimento di Chimica Organica “A. Mangini”, Universit ꢀa di Bologna, Viale Risorgimento 4, 40136 Bologna, Italy
S Supporting Information
b
ABSTRACT: Artemisinin or qinghaosu has now largely given way to the
more potent dihydroartemisinin (DHA, 1) and its derivatives in the
treatment of drug-resistant malaria, in combination with other classical
antimalarial drugs. DHA is obtained by NaBH reduction of artemisinin and
4
contains a stereochemically labile center at C-10, which provided two lactol
hemiacetal interconverting epimers, namely 1r and 1β. In the solid state,
the drug consists exclusively of the β-epimer; however, upon dissolution,
the two epimers equilibrate, reaching different solvent-dependent ratios
with different rates. Such equilibration also occurs in vivo, irrespective of the
isomeric purity at which the drug would have been administered. The aim of
this study was then to achieve an in-depth understanding of the kinetic
features of the R/β equilibration. To this purpose, free energy activation
q
barriers (ΔG ) of the interconversion were determined as a function of both
general and specific acid and base catalysts, ionic strength, and temperature in different solvents by dynamic HPLC (DHPLC). In
q
hydro-organic media, the dependence of ΔG on temperature led to the evaluation of the related enthalpic and entropic
contributions. Theoretical calculations suggested that the rate-determining step of the interconversion is not the ring-opening of the
cyclic hemiacetal but the previous reversible deprotonation of the individual epimers (base-catalyzed mechanism). The whole
findings may contribute to shed some light on the mechanism of action and/or bioavailability of the drug at the molecular level.
’
INTRODUCTION
also occurs in vivo, irrespective of the isomeric purity at which the
drug would have been administered. The intrinsic chemical
instability of DHA was subject of investigation by several
Malaria is a mosquito-borne infectious disease caused by a
eukaryotic protist of the genus Plasmodium. It is widespread in
tropical and subtropical regions; each year, there are approxi-
mately 350ꢀ500 million cases of malaria, killing between one
and three million people, the majority of whom are young
children in sub-Saharan Africa. In the past years, the combina-
tion of artemisinin derivatives with other antimalarial drugs
6bꢀe
groups
because it could prevent fulfillment of the interna-
tionally accepted stability requirements. Detailed studies on the
chemical and stereochemical behavior of DHA should therefore
have immediate relevance, inter alia, on all aspects related to drug
formulation and bioequivalence analysis.
1
7ꢀ16
In the past, numerous HPLC methods
have been developed
(
artemisinin-based combination therapies or ACTs) were devel-
for the analysis and plasma level monitoring of 1, but none of them
considers the effects of epimer equilibration on the analytical
response. Moreover, after a first study showing that interconversion
oped and are now the treatment of choice for the most lethal
2
,3
forms of malaria. Among such artemisinin derivatives, dihy-
droartemisinin (DHA, 1), which is also its main metabolite,
provides improved antimalarial potency.
stereochemically labile center at C-10, which provided two lactol
17
3
ꢀ5
of the two epimers occurred in a chromatographic time scale, no
DHA contains a
other related papers appeared in the literature afterward. Only
18
recently was this issue considered to provide improved and reliable
HPLC procedures suited for both separation and quantification of
the 1r and 1β epimers in solution and in complex mixtures, under
conditions of suppressed interconversion. In such a study, we
hemiacetal interconverting epimers, namely 1r and 1β
(
Chart 1). The R-epimer bears the hydroxyl group in the
equatorial position (absolute stereochemistry at C-10: R), and
6a
the β-epimer possesses an axial hydroxyl group. In the solid
state, DHA consists exclusively of the β-epimer; however, upon
dissolution, the two epimers equilibrate, reaching different
solvent-dependent ratios with different rates. Such equilibration
Received: December 2, 2010
Published: May 04, 2011
r 2011 American Chemical Society
4831
dx.doi.org/10.1021/jo102392p
|J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
Chart 1. Chemical Structures of Dihydroartemisinin (DHA, 1)
Epimers
(v/v/v) (hereafter denoted as WAM-solution) at controlled tem-
peratures in the range of 10ꢀ70 °C.
Solutions of 1 in the mobile phase were buffered or not as
needed to the apparent pH values (pH was measured in the ternary
hydro-organic mixtures, while the electrode system was calibrated
with aqueous standard buffers) ranging from 2.6 to 8.3 using
1
0 mM sodium phosphate. The measured first-order rate constants
for the forward (kβfR) and backward (kRfβ) interconversions
for details, see the Experimental Section) are reported in Table 1,
(
whereas the corresponding free energy activation barriers have
been calculated by the Eyring equation and are summarized in
Tables 1Sa,b of the Supporting Information.
identified some optimal conditions (such as stationary phase and
column temperature) able to minimize on-column epimerization
while achieving the best selectivity and efficiency of separation. This
paper, which is the second part of another focused on thermo-
dynamic considerations, deals with the kinetics of the 1βa1r
interconversion, with the aim of shedding light on the molecular
factors involved into the mechanism of action and/or toxicity of the
drug at molecular level.
In principle, such rate constants may be expressed by the
following general rate law, typical of hemiacetal breakdown
2
2
þ
ꢀ
ꢀ
kobs ¼ k0 þ kH½H ꢁ þ kOH½OH ꢁ þ kHAi½HAiꢁ þ kAi½A ꢁ
∑
∑
i
19
ð1Þ
where kobs generically represents the observed forward kβfR or
backward kRfβ first-order epimerization rate constants, k , kOH,
H
kHAi, and k are the second-order rate constants for the process
Ai
catalyzed by protons, hydroxide ions, general acids HA , and
i
’
RESULTS AND DISCUSSION
ꢀ
general bases A , whereas k expresses the possible first-order
i
0
17
The very first study showing that interconversion of the two
rate constant for the spontaneous epimerization in the given
medium. Because in the present case all the kinetic determina-
DHA epimers occurred on a chromatographic time scale (i.e., on-
column epimerization could be detected) was performed in 1986,
but since then, no other related papers have appeared in the
literature. In that study, the existence of the 1βa1r epimerization
was investigated at variable temperatures in aqueousꢀorganic
tions were performed by DHPLC, the k term will also include
0
the catalytic contribution coming from the stationary phase
(SPC) of the column here employed as the chromatographic
separating tool. Such a value for the same commercial column
(C18 symmetry, 3.5 μm; 75 ꢂ 4.6 mm i.d.) was already assessed
20
media (waterꢀethanol 55:45) by dynamic HPLC (DHPLC).
18a
However, the author reported a largely overestimated activation
in a previous study for both the forward βfR and backward
ꢀ
1
ꢀ3
energy barrier of 67.0 kcal mol for the interconversion of the two
DHA epimers, which cannot be related to dynamic events occur-
ring on the chromatographic time scale at room temperature. We
therefore recalculated by computer simulation the rate constants
from the published dynamic chromatograms and found much
lower energy barriers, actually consistent with those we recently
Rfβ epimerizations of DHA (SPC = 1.10 ꢂ 10 and 4.68 ꢂ
ꢀ4
ꢀ1
10
s , respectively). To get preliminary information on the
potential contributions of ionic strength to kobs we also carried
out further DHPLC determinations of the kβfR and kRfβ rate
constants in the same mobile phase but containing, in place of
phosphate, a catalytically inert salt (namely sodium chloride 1,
10, and 100 mM). The kinetic data obtained are collected in
Table 1. Comparison of the rate constants obtained with or
without phosphate buffer and salts (see Table 1) showed that
ionic strength has negligible effects on the reaction rate, within
the experimental error. This important result enabled us to
directly use the measured first-order kβfR and kβfR rate
constants (i.e., kobs in eq 1) to analyze the role of protons and
2
1
17
18a
reported within the same temperature range (see Figure 1S of
the Supporting Information for a collection of the relevant super-
imposed experimental and simulated chromatograms). Neverthe-
less, such kinetic information did not provide a satisfactory picture
of the process. In particular, the effects that a change of solvent, pH,
ionic strength, and temperature may generate on the rate of the
event were not investigated, and no information about the mechan-
ism could be obtained. Such a scenario prompted a thorough
kinetic investigation of the phenomena, with the aim of shedding
some light on the mechanism of action and/or bioavailability of the
drug at the molecular level.
Effect of pH, Ionic Strength, and Temperature on the Rate
Constants. Owing to the biological activity of 1, it is particularly
significant to investigate the kinetic features of the 1βa1r inter-
conversion in aqueous media, which can be assimilated to biological
fluids. Thus, we studied the epimerization in aqueousꢀorganic
mixtures as a function of both apparent pH and temperature and
measured the relevant forward and backward rate constants by
DHPLC (for details, see the Experimental Section). Chromato-
graphic conditions were selected in order to achieve the best
simulation of the relevant dynamic chromatograms with the lowest
ꢀ
4
2ꢀ
hydroxide ions, as well as that of H PO , H PO , and HPO4
3
4
2
species as promoters of the epimerization. As an example,
rateꢀpH profiles of the βfR epimerizations (i.e., the first-order
rate constants values plotted as a function of pH) at four
temperatures are reported in Figure 2. Inspection of the plots
clearly indicates that the process is effectively catalyzed by both
acid and basic species, reaching the minimum rate in the
restricted pH range from 4.0 to 5.5. According to the general
case expressed by eq 1, a more detailed form of dependence for
kobs, relative to the composition of the media employed as
reacting solvent in the explored pH range, may be conveniently
expressed by the following equation:
þ
ꢀ
kobs ¼ k0 þ kH½H ꢁ þ kOH½OH ꢁ þ kH3PO4 ½H3PO4ꢁ
18
associated error. All of the DHPLC experiments (examples of the
obtained chromatograms are shown in Figure 1, superimposed
with the relevant simulated traces) were carried out with a ternary
mobile phase consisting of waterꢀacetonitrileꢀmethanol 55:35:10
þ kH2PO4 ꢀ½H PO4 ꢁ þ kHPO4 ½HPO42ꢀ
ꢀ
2ꢀ
ꢁ
ð2Þ
2
Thus, in order to split out the contributions of each catalytic
species from the kobs data we performed new dedicated DHPLC
4
832
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
Figure 1. Superimposed experimental (red line) and simulated (blue line) dynamic chromatograms of 1 dissolved in mobile phase. Column: C18
ꢀ
1
symmetry, 3.5 μm (75 ꢂ 4.6 mm i.d.). Eluent: waterꢀacetonitrileꢀmethanol 55:35:10 (v/v/v). Flow rate: 1.0 mL min . Detection: UV at 214 nm. The
mobile phase was not buffered (top chromatograms) or buffered by 10 mM sodium phosphate to pH 2.55, 5.50, and 7.28 (bottom chromatograms).
determinations specifically created to quantify the input due to
autoprotolysis constants (i.e., changes of the constants pKa1,
pK , and pK ) were calculated according to the procedure
ꢀ
2ꢀ
the buffer’s components H PO , H PO , and HPO4 . For this
3
4
2
4
a2
w
22d
purpose, rate constant measurements at the selected temperature
described
for the transition from water to the binary
of 40 °C were repeated at seven established pHs (and then
aqueousꢀorganic solution waterꢀacetonitrile 65:35 (v/v), here-
after denoted as WA solution (that is to say, a mixture composi-
tion quite similar to that of the used ternary WAM-solution); (2)
the increment suffered by the pKa1 constant in pure acetonitrile
was roughly estimated by considering the typical changes re-
ported for the first stage of ionization of other polyprotic acids,
(i.e., tartaric and citric acid). From this, the relevant value in
WA solution was also estimated (a detailed description of the
whole procedure is reported in the Supporting Information). As a
reference, the same rate constants were also evaluated by
ꢀ
at seven different ratios [A ]/[HA ]) by progressively increa-
i
i
sing the buffer concentration inside the range 10ꢀ200 mM
(see Table 2S of the Supporting Information). The obtained
dynamic chromatograms, reported in Figure 2S of the Support-
ing Information, clearly evidence the existence of active general
catalysis by the buffer, which significantly increases on reduction
of pH. According to the mathematical methods extensively
22d
described within the Supporting Information, it was possible to
ꢀ
evaluate the second-order rate constants k
, and
H PO , kH PO
4
3
2
4
2
ꢀ
kHPO4 for the forward and backward epimerization of 1. It is
known that addition of organic modifiers to water solutions leads
to an increase of both pH of the media and ionization constants
resorting to pKa1, pKa2, and pK in pure water. All of the relevant
w
values were collected in Table 3S of the Supporting Information.
Information on these data allowed us to correct the experimental
kobs constants by subtracting from their values the calculable
buffer contributions at each relevant pH (i.e., the amounts
(
pK ) of neutral and anionic acid species, as well as also of the
a
solvent autoprotolysis constant value (pK ) related to the
w
ꢀ
ꢀ
2ꢀ
2ꢀ
22
kH3
4
2
4
[H PO4 ], and k
4
[HPO4 ]). The
employed waterꢀorganic mixture. While effects produced by
PO
[H3PO4], kH
PO
2
HPO
0
so elaborated new rate constants, k obs, deprived of any buffer
input, can afterward be expressed through the simplified eq 3:
addition of methanol or acetonitrile to water solutions have been
22a,b
extensively investigated,
no equivalent data are available for
ternary mixtures in the literature. In addition, whereas changes
of ionization constant for the dihydrogen phosphate anion
0
þ
ꢀ
k
¼ k0 þ kH½H ꢁ þ kOH½OH ꢁ
ð3Þ
obs
(
i.e., pK ) can be estimated according to detailed procedures
The unknown kinetic quantities k , k , and k for the
a2
0
H
OH
reported for the limited case of binary mixtures involving either
forward and backward epimerization of 1 have then been
2
2c
22d,e
methanol or acetonitrile
as the cosolvent, no equivalent
obtained (see Table 3S and Figure 3S of the Supporting
0
information may be found about the first stage of ionization of
Information) by fitting the scatter plot data pH vs k
to eq 3
obs
the orthophosphoric acid (i.e., pK ). Thus, to perform evalua-
according to a nonlinear regression procedure. The reliability of
such a procedure was supported by the achievement of really
good correlation coefficients (from the treatment with water
a1
tions of the above-cited second-order rate constants in WAM-
solution, we resorted to the following approximations: (1)
all changes of the involved solute ionization and solvent
2
data, nine data points: R = 0.985 for the βfR epimerization,
4
833
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
Table 1. Pseudo-First-Order Epimerization Rate Constants Measured at Different Temperatures and pHs in Either Buffered
Ternary Aqueous Mixtures or Unbuffered Solutions of Different Ionic Strength
pH
T (°C)
2.55
3.01
3.80
3.92
5.50
6.00
6.87
7.28
2.5
8.25
8.1
4
ꢀ1
a
a
a
a
c
k
Rfβ ( ꢂ 10 s
)
phosphate buffer (10 mM)
10
2.1
1.1
0.8
0.3
0.5
1.3
2.3
a
a
a
a
c
20
26
30
35
40
45
50
60
70
4.8
8.9
2.7
1.7
0.7
1.2
6.5
21.8
a
4.5
2.6
b
10.8
16.3
23.4
44.4
58.5
6.8
3.5
1.8
4.6
6.7
b
12.3
15.9
26.8
5.1
d
7.4
5.7
9.2
7.3
17.8
55.8
29.4
61.8
81.4
b
10.6
15.3
32.4
a
a
a
a
43.0
10.2
23.2
44.9
12.4
162.3
NaCl (1 mM)
20
20
20
30
3.2
2.5
2.7
5.6
9.4
13.8
2.8
4.7
NaCl (10 mM)
NaCl (100 mM)
no addition of buffers and salts
3
4
5
0
4
ꢀ1
a
a
a
a
c
k
βfR ( ꢂ 10 s
)
phosphate buffer (10 mM)
10
4.2
2.2
1.5
0.5
0.9
4.4
12.0
33.5
a
a
a
a
c
20
26
30
35
40
45
50
60
70
9.9
5.5
3.3
1.4
2.3
11.8
a
18.0
22.1
33.4
49.9
91.9
9.2
5.2
b
a
14.1
25.7
33.4
56.3
90.9
7.1
3.6
9.5
14.6
39.8
125.4
b
b
10.5
15.4
22.8
33.4
73.3
d
11.9
19.9
15.5
27.0
62.5
173.9
380.2
a
a
a
127.3
21.9
51.0
99.0
136.8
NaCl (1 mM)
20
20
20
30
6.3
NaCl (10 mM)
5.2
NaCl (100 mM)
5.4
no addition of buffers and salts
11.8
19.8
29.6
3
4
5
0
a
d
q
b
q
c
q
Data extrapolated by correlations 1/T vs ΔG /T. Data interpolated by correlations 1/T vs ΔG /T. Data interpolated by correlations pH vs ΔG .
Data extrapolated by correlations pH vs ΔG . Errors associated to the kinetic data from DHPLC were e12%.
q
2
ꢀ4 ꢀ1
ꢀ4 ꢀ1
R = 0.987 for the Rfβ interconversion, respectively; from the
term and 5.50 ꢂ 10
s
from water data, 4.7ꢂ 10
s
from
2
treatment with WA solution data, nine data points: R = 0.988 for
the βfR epimerization, R = 0.991 for the Rfβ interconver-
sion, respectively).
From the above results, it is evident that, irrespective of the
considered solvent reference (i.e., pure water or WA solution),
the specific base-catalysis is much more efficient than the acid-
catalysis. Furthermore, the specific base-catalysis operated by
OH also appeared to be much more effective than the general
catalysis performed by the buffer anions. A quite similar behavior
was already reported for the breakdown of several aryl hemiketals
of R-bromoacetophenone. On the contrary, the specific acid-
catalysis by H was shown to be quite similarly efficient to the
corresponding general acid-catalysis by H PO . About the con-
stant term k , the fitting procedure afforded values very close to
those expected for the catalytic contributions coming from the
WA solution data for the backward k term by the fitting
0
2
ꢀ3
procedure against the SPC quantities 1.10 ꢂ 10 and 4.68 ꢂ
ꢀ
4 ꢀ1
10
s
previously obtained as effects of the stationary phase for
18
the same column, respectively), so that such an input was
assumed to have, in its essence, just this origin. To get a more
detailed picture of the catalytic contributions given to k by the
obs
þ
ꢀ
ꢀ
4
2ꢀ
single H , OH , H PO , H PO , and HPO4 species as a
3
4
2
ꢀ
function of their efficiency (i.e., the related value of the second-
order rate constant) and concentration, we calculated for the
explicit temperature of 40 °C their individual contribution at
each established pH in the 2ꢀ8 range by splitting eq 2 into the
respective components. The results obtained with reference to
thermodynamic data coming from both pure water and WA
solution are collected in Table 4S of the Supporting Information
and plotted in Figures 3 and 4 (only data related to the 1βf1r
interconversion). Obviously, the inevitable errors, in absolute
sense, involved by the above exposed approximations will be
23a
þ
3
4
0
ꢀ
3
ꢀ1
chromatographic stationary phase (1.2 ꢂ 10
s
from water
ꢀ3
ꢀ1
data, 1.0 ꢂ 10
s
from WA solution data for the forward k0
4
834
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
Figure 2. RateꢀpH profiles of the 1βf1r epimerization as a function of temperature, in the 10ꢀ50 °C range.
Figure 3. Calculated specific (SpecificC) and general (GeneralC) catalytic contributions to the kobs pseudo-first-order rate constant at 40 °C for the
βf1r epimerization as a function of pH in waterꢀacetonitrileꢀmethanol 55:35:10 (v/v/v) solutions buffered by 10 mM sodium phosphate salts.
With SPC and TotalC are also reported (i) the constant catalytic contribution afforded to kobs by the chromatographic stationary phase and (ii) the sum
1
w
of the three contributions SpecificC, GeneralC, and SPC at each considered pH. (a) Data from pKa1, pKa2 and pK in water. (b) Data from pKa1, pKa2,
and pK in WAꢀsolution.
w
reflected in a systematic, although moderate, way on the whole
explored range of pH; nevertheless, the so-evaluated buffer
effects, translated to the real case of biological fluids where much
more active species are certainly present, should retain an
important and explanatory relevance.
Under this value the trend is inverted. (ii) In the 4.5ꢀ6.5 pH range,
ꢀ
H PO plays the prevalent catalytic role, with minimal additive
2
4
þ
2ꢀ
contributions coming from the acidic H and HPO
and basic
4
ꢀ
OH species in the more restricted 5ꢀ6 pH range. (iii) In the pH
2ꢀ
interval from 6.5 to 8, the double-charged species HPO4 does not
Inspection of the results coming from the medium treated as WA-
show any input, while a residual provision of catalysis is again
ꢀ
solutions reveals the following points: (i) Below pH 4.5, a greater
afforded by the H PO4 ions. Instead, a completely dominant and
2
þ
ꢀ
catalytic contribution comes from both H and H PO than from
2
growing catalytic effect is displayed by OH ions. One conclusion
3
4
ꢀ
4
H PO . In particular, the inputs from the two acidic species show a
that can be made is that the base-promoted epimerization of 1 is
prevalence of the general acid on the specific one up to pH 2.5.
largely preferred over the acid-promoted epimerization, the first
4
835
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
Figure 4. Calculated contributions of single catalytic species to the kobs pseudo-first-order rate constant at 40 °C for the 1rf1β epimerization as a
function of pH in waterꢀacetonitrileꢀmethanol 55:35:10 (v/v/v) solutions buffered by 10 mM sodium phosphate salts. Top view: logarithms of the
absolute contributions to k vs pH. Bottom view: percent contributions to k vs pH. (a) Data from pK , pK , and pK in water. (b) Data from pK ,
obs
obs
a1
a2
w
a1
pK , and pK in WA solution.
a2
w
being the dominant mechanism at pH > 4.5 (in this context, it was
This evidence strongly suggests that the involvement of basic
ꢀ8
assumed that, because of the low value of 6.2 ꢂ 10 M of their
species effectively makes the hemiacetal ring in 1r and 1β ready
to undergo an almost spontaneous breakdown and that the rate-
limiting step of such a kinetic pathway should be the previous
equilibrium of proton transfer from the substrate to the basic
catalyst (step [1]a of Scheme 1), rather then the ring-opening in
ꢀ
ionization constant, the ions H PO act as basic species).
2
4
Molecular Modeling Calculations. To find a rational expla-
nation of the above results we resorted to the molecular modeling
19
calculations already performed (see also the Experimental
ꢀ
ꢀ
ꢀ
ꢀ
Section) on the structure of 1r and 1β anions, as well as on
that of the conjugate acids 1r and 1β in water, acetonitrile, and
hexane. In all cases, their resulting geometries showed a varying,
substantial degree of opening of the hemiacetal cycle. In parti-
cular, it appeared that the extent of breakdown becomes more
pronounced moving from acetonitrile or water to hexane and
even more in the gas phase (see Figure 5). Notably, the extent of
bond cleavage of the CꢀO bond in R position to the hemiacetal
the anionic intermediates 1r and 1β (step [1]b).
23
Hemiacetal Ring-OpeningMechanism. As reported elsewhere,
we faced a scenario typical of hemiacetals and hemiketals derived from
strongly acid alcohols, whose pK values approach or are smaller
a
than ∼13. The conjugate bases resulting by deprotonation of these
latter species manifest, in fact, a resolute aptitude to be very good
leaving groups. Consequently, in the base-catalyzed mechanism, as
soon as they are formed, they may undergo breakdown faster than
they can reprotonate and revert to the original neutral form. Some
relevant examples of hemiketals deriving by parent alcohols with pKa
values smaller or close to the above limit are given by the 2,2,2-
trifluoroethyl, 4-methoxyphenol, and 3-chlorophenol hemiketals of
ꢀ
carbon is strongly progressed in the case of 1r anion (70% in
gas phase and 68% in water and acetonitrile), while it is less
ꢀ
advanced for 1β (51% in gas phase and 29% in water and
acetonitrile). This is particularly evident when looking at the
2
3a
CꢀO atomic distances of the breaking bond in the optimized
R-bromoacetophenone (pK of the parent alcohols: 12.37, 10.20,
a
ꢀ
ꢀ
24
structures of 1r and 1β and in the opened structure 3 (see
Scheme 1), where the CꢀO bond is completely broken
and 9.02). By using two commercial software programs, we
estimated for the alcoholic function resulting from the breakdown
of the two epimers of 1 (i.e., the conjugate acid 3 of the anionic
(Figure 5).
4
836
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
Figure 5. Extent of bond cleavage of the CꢀO bond, in R position to the hemiacetal carbon of 1, estimated by DFT calculations (for details, see the
Experimental Section). Interatomic distances are expressed in nm.
Scheme 1. Base- (Pathway [1]) and Acid-Catalyzed (Pathway [2]) Breakdown of 1r and 1β
2
4a
intermediate 2 reported in Scheme 1) the pK values of 11.32 and
1
1Sa of the Supporting Information). At the same time, according to
the superimposed existence of different catalytic mechanisms and
species operating in the explored pH range, the calculated activation
a
24b
2.42. Apart from the small difference between these values, they
were always smaller than 13. Thus, this finding also gave further
support to the hypothesis that, in the base-catalyzed epimerization
of 1, deprotonation is the rate-limiting step. In tight analogy with the
results found in the absence of buffers (see Table 1Sb of the
Supporting Information), the van’t Hoff analysis carried out on the
activation barriers related to the first-order rate constants given in
Table 1 indicated that the time-limiting step of the epimerization of
q
q
enthalpies ΔH
and ΔH
showed varying values as a
βꢀR
Rꢀβ
function of pH. These, in fact, gave place to bell-shaped pro-
files that achieved the maximum in the restricted 5.0ꢀ6.5 pH range
q
(see Figure 6). Interestingly, upon pH variation, more negative ΔS
q
values correspond to smaller values of ΔH . As a consequence, at a
given temperature, these opposite trends restrain the difference
q
1
is only slightly dependent on the temperature; in fact, at all the
between the maximum and minimum value that ΔG may assume
considered pH values, the calculated variations of activation entropy,
in response to pH changes.
To conclude the kinetic study on the epimerization of 1, we
performed a van’t Hoff analysis on the activation barriers related
q
q
i.e., ΔS
and ΔS
, have small negative values (ꢀ26 ( 3
Rꢀβ
βꢀR
and ꢀ23 ( 3 e.u. are the averaged amounts, respectively; see Table
4
837
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
determined in the ternary mixture. By inspection of the results
collected in Table 5S of the Supporting Information, we con-
cluded that in all the considered physiological media, with the
exception of saliva and urine, the calculated half-life times of 1
were smaller than 10 min. This clearly means that, from a kinetic
point of view, a significant presence of both epimers, correspond-
ing to the equilibrium ratio KR/β, would be found after about 1 h
in such biological systems, irrespective of the isomeric purity at
which 1 would have been administered. From a thermodynamic
point of view, however, it should be also taken into account that
the relative abundance of 1r and 1β could be influenced by the
presence of species (typically, globular proteins) able to form
complexes of different stability, thus shifting the 1r/1β equilib-
rium position.
ꢀ
1
Figure 6. Activation enthalpies (kcal mol ) of the 1βf1r (top curve)
’
CONCLUSIONS
and 1rf1β epimerizations (bottom curve) expressed as a function of
pH in waterꢀacetonitrileꢀmethanol 55:35:10 (v/v/v) solutions buf-
fered by 10 mM sodium phosphate.
The present study was conceived to achieve information about
kinetic and mechanistic features of the 1r/1β equilibration by
both experimental and theoretical approaches. As expected, the
epimerization rates were found to be sensitive to both acids and
bases. However, in the solvent systems tested, the specific base-
catalysis was demonstrated to be much more effective than the
specific acid-catalysis, as well as more influencing than the
corresponding general catalysis in the analyzed buffered ternary
to the first-order rate constants measured at each considered pH
value in the 2ꢀ8 range. Considering the results on both forward
and backward epimerization of 1 (see Table 1Sa of the Support-
ing Information), it was pointed out that, in the examined ternary
hydro-organic mixture, the enthalpic activation barriers for the
rf1β isomerization are in all cases lower than those for the
opposite conversion (i.e., ΔΔH = ΔH
mixture. Instead, acid-catalysis by both specific and general
species (H and H
1
þ
q
q
q
3
PO
4
, respectively) was shown to be of quite
ꢀ ΔH
< 0),
βꢀR
Rꢀβ
similar efficiency and dominant at pH < 4.5. In such a system, the
epimerization rates were measured and analyzed as a function of
ionic strength, pH, and temperature, and it was found that a base-
catalyzed mechanism was responsible for the epimerization in
quantitative terms at pH values greater than 4.5. Moreover,
theoretical calculations suggested that the rate-limiting step of
the process was not the ring-opening of the cyclic hemiacetal but
the previous deprotonation equilibrium of the individual epi-
mers. By dividing the entropic and enthalpic contributions to the
free energy activation barriers calculated as function of pH for
both the forward and backward epimerizations, it was also found
that the observed R/β concentration ratio is entropic in origin,
according to a differential solvation of the transition states in the
rate-limiting step of proton transfer. This, in turn, can be related
and this is in contrast with the corresponding experimental
ΔΔG values that, in the range of the explored temperatures,
q
are always greater than zero (that is, the related KR/β equilibrium
constants are >1). However, the right correspondence is
promptly restored when entropic contributions are taken into
account. Such results suggest that the experimental relative
abundance of the 1r and 1β epimers in the ternary medium is
essentially entropic in origin or else the K constant would be
R/β
smaller than one by the effect of the enthalpic term only. This is
consistent with a differential solvation of the transition states
developed during the rate-limiting step of deprotonation, which
may be predicted to be significantly more effective than that
involving the related ground states. Such transition states feature
advanced degrees of proton transfer (more than 80%) toward the
19
2
3a
to the different acidity already found for 2r and 2β.
catalyzing base (i.e., they have a strong anionic character), in
an extent modulated by the hemiacetal acidity. Therefore, the
We believe that the present results on the stereolability of 1
may contribute to shed some light on the mechanism of action
and/or bioavailability of the drug at molecular level and represent
a promising starting point for future studies and rational design of
semisynthetic artemisinin derivatives.
19
different acidity that we found for 1r and 1β may easily
accounts for the observed differential effect on entropy; this
suggests the formation of stronger interactions of the solvent
molecules (which, then, reduce their entropy more distinctly)
with the transition state related to 1r compared to 1β, because
the more acidic epimer should clearly undergo a more advanced
degree of proton transfer.
Half-Life Times Prediction in Physiological Fluids. With the
above kinetic information on the βaR interconversion of 1 in
hand, it was possible to calculate approximate half-life times (t1
’
EXPERIMENTAL SECTION
Chromatography. Analytical liquid chromatography was per-
formed using a Waters model 2695 HPLC separation module coupled
with a Waters 996 photodiode array detector. Chromatographic data
were collected and processed using Empower2 software (Waters).
Variable-temperature HPLC (DHPLC) was performed by using a
laboratory-made thermally insulated container cooled by the expansion
)
/2
for the 1r and 1β species allowed to equilibrate within physio-
logical fluids found in different human organs or tissues at 40 °C
(
see Table 4S of the Supporting Information). Because in these
2 2
of liquid carbon dioxide (CO ). Flow of liquid CO and column
þ
ꢀ
estimations we considered H and OH ions as the exclusive
catalytic species, the calculated t1/2 times have to be considered as
upper limit values, which could therefore be further reduced by
general acid and basic species present in such media. As an
approximation, it was assumed that, in such fluids, the second-
order rate constants kH and kOH at 40 °C were the same as
temperature were regulated by a solenoid valve, thermocouple, and
electric controller. Temperature variations after thermal equilibration
were within (0.2 °C.
Normal-phase HPLC for the classic batchwise approach was per-
formed on a Reprosil Si 120, 3 μm (150 ꢂ 4.6 mm i.d.) column with a
mobile phase made up of n-hexaneꢀethanol 97:3 (v/v), delivered at
4
838
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
ꢀ
1
1
.0 mL min , and UV detection at 214 nm (T = 25 °C). Reversed-phase
Universit ꢀa di Roma, Italy (funds for selected research topics
2008ꢀ2010).
HPLC was performed on a C18 symmetry, 3.5 μm (100 ꢂ 4.6 mm i.d.)
column, with a mobile phase composed of waterꢀacetonitrileꢀ
methanol 55:35:10 (v/v/v) buffered to apparent pH 5.6 using 10 mM
sodium phosphate, delivered at 1.0 mL min , and UV detection at
14 nm (T = 5 °C).
ꢀ
1
’ REFERENCES
(1) Guidelines for the treatment of malaria. World Health Organiza-
2
tion, 2009.
All the DHPLC experiments were performed on a C18 symmetry,
.5 μm (75 ꢂ 4.6 mm i.d.) column, with a ternary hydro-organic mobile
(2) O’Neill, P. M. Expert Opin. Invest. Drugs 2005, 14, 1117–1128.
3
ꢀ1
(3) (a) O’Neill, P. M.; Posner, G. H. J. Med. Chem. 2004,
7, 2945–2964. (b) Bez, G.; Kalita, B.; Sarmah, P.; Barua, N. C.; Dutta,
phase delivered at 1.0 mL min and UV detection at 214 nm. For the
study of the effects of pH and temperature on the rate constants, a
ternary mobile phase consisting of waterꢀacetonitrileꢀmethanol
4
D. K. Curr. Org. Chem. 2003, 7, 1231–1255.
4) Posner, G. H.; Paik, I.-H.; Chang, W.; Borstnik, K.; Sinishtaj, S.;
Rosenthal, A. S.; Shapiro, T. A. J. Med. Chem. 2007, 50, 2516–2519.
5) Li, Q. G.; Peggins, J. O.; Fleckenstein, L. L.; Masonic, K.; Heiffer,
M. H.; Brewer, T. G. J. Pharm. Pharmacol. 1998, 50, 173–182.
6) (a) Pathak, A. K.; Jain, D. C.; Sharma, R. P. Indian J. Chem., Sect.
(
5
1
5:35:10 (v/v/v), at controlled temperatures in the range of
ꢀ
1
0ꢀ70 °C was tested. Solutions of 1 (0.25ꢀ1.0 mg mL ) in the
(
mobile phase were buffered to apparent pH values ranging from 2.6 to
8
.3 using 10 mM sodium phosphate. Aliquots of 10ꢀ20 μL were
(
injected. Increasing concentrations of sodium phosphate buffer inside
the range 10ꢀ200 mM were used in the study addressed to quantify the
catalytic role of each buffer’s component (see Table 2S and Figure 2S of
the Supporting Information for details about the employed concen-
trations). Sodium chloride (1, 10, and 100 mM) replaced sodium
phosphate in the ternary mobile phase waterꢀacetonitrileꢀmethanol
B: Org. Chem. Incl. Med. Chem. 1995, 34, 992–993. (b) Haynes, R. K.;
Chan, H.-W.; Lung, C.-M.; Ng, N.-C.; Wong, H.-N.; Shek, L. Y.;
Williams, I. D.; Cartwright, A.; Gomes, M. F. Chem. Med. Chem. 2007,
2, 1448–1463. (c) Dhooghe, L.; Van Miert, S.; Jansen, H.; Vlietinck, A.;
Pieters, L. Pharmazie 2007, 62, 900–901. (d) Jansen, F. H.; Soomro,
S. A. Curr. Med. Chem. 2007, 14, 3243–3259. (e) Haynes, R. K. Chem.
Med. Chem. 2006, 6, 509–537.
5
5:35:10 (v/v/v) in the study of the role of ionic strength.
(7) Zhou, Z. M.; Anders, J. C.; Chung, H.; Theoharides, A. D.
Simulation of Dynamic Chromatograms. Simulations of
experimental dynamic chromatograms were performed by using the
Auto DHPLC y2k laboratory-made computer program, which imple-
ments both stochastic and theoretical plate models and may take into
account all types of first-order interconversions as well as tailing effects.
In the present paper, all simulations were carried out by using the
stochastic model and taking tailing effects into consideration. The rate
constants in the mobile phase were properly set in order to obtain a 1r/
β ratio consistent with the thermodynamic KR/β ratio experimentally
measured in the same media by H NMR, according to a procedure
reported elsewhere.
Molecular Modeling Calculations. Geometries of 1r, 1β, 1r ,
J. Chromatogr. 1987, 414, 77–90.
(8) Karbwang, J.; Na-Bangchang, K.; Molunto, P.; Banmairuroi, V.;
Congpuong, K. J. Chromatogr. B 1997, 690, 259–265.
21
20
(9) (a) Batty, K. T.; Davies, T. M. E.; Thu, L. T.; Binh, T. Q.; Anh,
T. K.; Ilett, K. F. J. Chromatogr. B 1996, 677, 345–350. (b) Batty, K. T.;
Ilett, K. F.; Davis, T. M. E. J. Pharm. Pharmacol. 1996, 48, 22–26.
(c) Kotecka, B. M.; Rieckmann, K. H.; Davis, T. M. E.; Batty, K. T.; Ilett,
K. F. Acta Trop. 2003, 87, 371–375. (d) Batty, K. T.; Ilett, K. F.; Davis,
T. M. E. Br. J. Clin. Pharmacol. 2004, 57, 529–533.
1
1
(10) (a) Navaratnam, V.; Mordi, M. N.; Mansor, S. M.; Chin, L. K.;
18
Asokan, M.; Nair, N. K. J. Chromatogr. B 1995, 669, 289–294.
(b) Navaratnam, V.; Mordi, M. N.; Mansor, S. M. J. Chromatogr. B
1997, 692, 157–162. (c) Lai, C.-S.; Nair, N. K.; Mansor, S. M.; Olliaro,
P. L.; Navaratnam, V. J. Chromatogr. B 2007, 857, 308–314.
(11) (a) Sandrenan, N.; Sioufi, A.; Godbillon, J.; Netterb, C.;
Danker, M.; van Valkenburg, C. J. Chromatogr. B 1997, 691, 145–153.
ꢀ
ꢀ
1
β , and 3 were optimized at the SCF level by DFT calculations run
with the Amsterdam Density Functional (ADF) package v. 2007.01. The
employed method was the GGA-BLYP/DZP medium core basis set. All
of the relative solvation energies in hexane, acetonitrile, and water were
computed with the same program by the COnductor like Screening
MOdel (COSMO), with the cavity defined according to the Solvent
Excluding Surface (SES) algorithm.
(
b) Souppart, C.; Gauducheau, N.; Sandrenan, N.; Richard, F.
J. Chromatogr. B 2002, 774, 195–203.
12) Avery, B. A.; Venkatesh, K. K.; Avery, M. A. J. Chromatogr. B
999, 730, 71–80.
13) Ortelli, D.; Rudaz, S.; Cognard, E.; Veuthey, J.-L. Chromato-
graphia 2000, 52, 445–450.
14) (a) Sabarinath, S.; Rajanikanth, M.; Madhusudanan, K. P.;
(
1
(
’
ASSOCIATED CONTENT
(
S
Supporting Information. Activation parameters for the
b
Gupta, R. C. J. Mass Spectrom. 2003, 38, 732–742. (b) Rajanikanth,
M.; Madhusudanan, K. P.; Gupta, R. C. Biomed. Chromatogr. 2003, 17,
forward and backward first-order epimerization of 1. Second-
order rate constants based on specific and general acid and basic
catalysis. Contributions to the forward and backward pseudo-
first-order epimerization rate constants. Dynamic chromato-
grams digitalized by ref 17. Cartesian coordinates of 1r, 1β,
440–446.
(15) Naik, H.; Murry, D. J.; Kirsch, L. E.; Fleckenstein, L.
J. Chromatogr. B 2005, 816, 233–242.
(16) Liu, Y.; Zeng, X.; Deng, Y.; Wang, L.; Feng, Y.; Yang, L.; Zhou,
D. J. Chromatogr. B 2009, 877, 465–470.
ꢀ
ꢀ
1
r , 1β , and 3. This material is available free of charge via the
Internet at http://pubs.acs.org.
(17) Shishan, Z. Chromatographia 1986, 22, 77–80.
(18) (a) Cabri, W.; Ciogli, A.; D’Acquarica, I.; Di Mattia, M.;
Galletti, B.; Gasparrini, F.; Giorgi, F.; Lalli, S.; Pierini, M.; Simone, P.
J. Chromatogr. B 2008, 875, 180–191. (b) D’Acquarica, I.; Gasparrini, F.;
Kotoni, D.; Pierini, M.; Villani, C.; Cabri, W.; Di Mattia, M.; Giorgi, F.
Molecules 2010, 15, 1309–1323.
’
AUTHOR INFORMATION
Corresponding Author
*
E-mail: (F.G.) francesco.gasparrini@uniroma1.it; (M.P.) marco.
(19) Cabri, W.; D’Acquarica, I.; Simone, P.; Di Iorio, M.; Di Mattia,
pierini@uniroma1.it.
M.; Gasparrini, F.; Giorgi, F.; Mazzanti, A.; Pierini, M.; Quaglia, M.;
Villani, C. J. Org. Chem. 2011, 76, 1751–1758.
(
20) (a) Keller, R. A.; Giddings, J. C. J. Chromatogr. 1960,
, 205–220. (b) Kramer, R. J. Chromatogr. 1975, 107, 241–252.
c) Schurig, V.; B €u rkle, W. J. Am. Chem. Soc. 1982, 104, 7573–7580.
(d) B €u rkle, W.; Karfunkel, H.; Schurig, V. J. Chromatogr. 1984,
288, 1–14. (e) Veciana, J.; Crespo, M. I. Angew. Chem., Int. Ed. Engl.
’
ACKNOWLEDGMENT
3
(
We acknowledge financial support from FIRB [research
program: Ricerca e Sviluppo del Farmaco (CHEM-PROFAR-
MA-NET), grant no. RBPR05NWWC_003] and from Sapienza
4
839
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840

The Journal of Organic Chemistry
ARTICLE
1
991, 30, 74ꢀ77. (f) Jung, M. QCPE Bull. 1992, 12, 52. (g) Cabrera, K.;
Jung, M.; Fluck, M.; Schurig, V. J. Chromatogr. A 1996, 731, 315–321.
(h) Oxelbark, J.; Allenmark, S. J. Org. Chem. 1999, 64, 1483–1486.
(i) Trapp, O.; Schoetz, G.; Schurig, V. Chirality 2001, 13, 403ꢀ414.
(j) Wolf, C. Chem. Soc. Rev. 2005, 34, 595–608. (k) Trapp, O. Anal.
Chem. 2006, 78, 189–198. (l) D’Acquarica, I.; Gasparrini, F.; Pierini, M.;
Villani, C.; Zappia, G. J. Sep. Sci. 2006, 29, 1508–1516.
(21) (a) Alcaro, S.; Casarini, D.; Gasparrini, F.; Lunazzi, L.; Villani,
C. J. Org. Chem. 1995, 60, 5515–5519. (b) Gasparrini, F.; Lunazzi, L.;
Mazzanti, A.; Pierini, M.; Pietrusiewicz, K. M.; Villani, C. J. Am. Chem.
Soc. 2000, 122, 4776–4780. (c) Dell’Erba, C.; Gasparrini, F.; Grilli, S.;
Lunazzi, L.; Mazzanti, A.; Novi, M.; Pierini, M.; Tavani, C.; Villani, C.
J. Org. Chem. 2002, 67, 1663–1668. (d) Gasparrini, F.; Grilli, S.; Leardini,
R.; Lunazzi, L.; Mazzanti, A.; Nanni, D.; Pierini, M.; Pinamonti, M.
J. Org. Chem. 2002, 67, 3089–3095. (e) Dalla Cort, A.; Gasparrini, F.;
Lunazzi, L.; Mandolini, L.; Mazzanti, A.; Pasquini, C.; Pierini, M.;
Rompietti, R.; Schiaffino, L. J. Org. Chem. 2005, 70, 8877–8883.
(
f) Cirilli, R.; Ferretti, R.; La Torre, F.; Secci, D.; Bolasco, A.; Carradori,
S.; Pierini, M. J. Chromatogr. A 2007, 1172, 160–169.
22) (a) Subirats, X.; Ros ꢁe s, M.; Bosch, E. Sep. Purif. Rev. 2007,
6, 231–255. (b) Ros ꢁe s, M.; Bosch, E. J. Chromatogr. A 2002, 982, 1–30.
c) Bosch, E.; Bou, P.; Allemann, H.; Ros ꢁe s, M. Anal. Chem. 1996,
(
3
(
68, 3651–3657. (d) Bosch, E.; Espinosa, S.; Ros ꢁe s, M. J. Chromatogr. A
1998, 824, 137–146. (e) Espinosa, S.; Bosch, E.; Ros ꢁe s, M. Anal. Chem.
2000, 72, 5193–5200.
(23) (a) McClelland, R. A.; Kanagasabapathy, V. M.; Mathivanan, N.
Can. J. Chem. 1991, 69, 2084–2093. (b) Sorensen, P. E.; Jencks, W. P.
J. Am. Chem. Soc. 1987, 109, 4675–4690. (c) Sorensen, P. E.; Pedersen,
K. J.; Kanagasabapathy, V. M.; McClelland, R. A. J. Am. Chem. Soc. 1988,
1
10, 5118–5123.
24) (a) ACD/Labs PackageV. 6.00, (pKa DB Module), 90 Adelaide
Street, West Toronto, Ontario M5H 3V9, Canada. (b) Marvin 4.1.1.3,
007, ChemAxon (http://www.chemaxon.com). These packages were
used to draw and calculate pK values of hemiacetals.
(
2
a
4
840
dx.doi.org/10.1021/jo102392p |J. Org. Chem. 2011, 76, 4831–4840