1066 Biomacromolecules 2004, 5, 1066-1077

Molecular Modeling and Simulation of Mycobacterium

tuberculosis Cell Wall Permeability
Xuan Hong and A. J. Hopfinger*
Laboratory of Molecular Modeling and Design (MC 781), College of Pharmacy, The University of Illinois at
Chicago, 833 South Wood Street, Chicago, Illinois 60612-7231
Received December 8, 2003; Revised Manuscript Received February 9, 2004

The low permeability of the mycobacterial cell wall is thought to contribute to the intrinsic drug resistance
of mycobacteria. In this study, the permeability of the Mycobacterium tuberculosis cell wall is studied by
computer simulation. Thirteen known drugs with diverse chemical structures were modeled as solutes
undergoing transport across a model for the M. tuberculosis cell wall. The properties of the solute-membrane
complexes were investigated by means of molecular dynamics simulation, especially the diffusion coefficients
of the solute molecules inside the cell wall. The molecular shape of the solute was found to be an important
factor for permeation through the M. tuberculosis cell wall. Predominant lateral diffusion within, as opposed
to transverse diffusion across, the membrane/cell wall system was observed for some solutes. The extent of
lateral diffusion relative to transverse diffusion of a solute within a biological cell membrane may be an
important finding with respect to absorption distribution, metabolism, elimination, and toxicity properties
of drug candidates. Molecular similarity measures among the solutes were computed, and the results suggest
that compounds having high molecular similarity will display similar transport behavior in a common
membrane/cell wall environment. In addition, the diffusion coefficients of the solute molecules across the
M. tuberculosis cell wall model were compared to those across the monolayers of dipalmitoylphosphati-
dylethanolamine and dimyristoylphosphatidylcholine, are two common phospholipids in bacterial and animal
membranes. The differences among these three groups of diffusion coefficients were observed and analyzed.

Introduction Mycobacterial mycolic acids have several distinctive

Tuberculosis (TB) kills approximately 2 million people
each year.1 The global incidence of TB is still increasing at
approximately 0.4% per year and much faster in some areas.2
The devastating impact of this disease is attributed to three
factors: the breakdown in health services, the co-infection
with HIV/AIDS, and the emergence of multidrug-resistant
TB.1
Mycobacteria infections are, in general, difficult to treat
because mycobacteria are resistant to the majority of common
antibiotics and chemotherapeutic agents.3,4 An important
factor causing this resistance is the barrier imposed by the
mycobacterial cell wall.5-10 The mycobacterial cell wall is
extraordinarily thick and tight, having two main compo-
nents: (1) characteristic long chain fatty acids, mycolic acids,
and (2) unique polysaccharides, arabinogalactan (AG).11-13
These two constituents are covalently linked together by ester
bonds. The mycolyl-AG complex is then attached to
peptidoglycan, a porous layer between the cell wall and the
plasma membrane, and form the mycolyl-AG peptidoglycan
complex. The mycobacterial cell wall also contains many
“free” lipid species, the so-called extractable lipids, that are
not covalently linked to the AG-peptidoglycan complex and
are solvent-extractable. The free lipids include glycolipids,
phenolic glycolipids, glycopeptidolipids, and other chemical
species.11-17
* Corresponding author. Telephone: 312.996.4816. Fax: 312.413.3479.
E-mail: hopfingr@uic.edu.
10.1021/bm0345155 CCC: $27.50
Published on Web 04/06/2004
features compared to most fatty acids: (1) they are long-
chain molecules having a long branch, mero chain, of 40-
60 carbons and a short branch, R branch, of typically 24
carbons; and (2) in addition to their extraordinary lengths,
mycolic acids contain very few double bonds or cyclopropane
groups. The short branch, R branch, is totally saturated, and
the long branch, mero chain, only has two positions that are
observed to be occupied by functional groups. Of these two
positions, the proximal position (nearer the β-hydroxy acid)
contains exclusively cis- or trans-olefin or cyclopropane,
while the distal position may be the same as the proximal
position or may contain one of a variety of oxygen moieties
such as R-methyl ketone, R-methyl methyl ether, methyl-
branched ester, or R-methyl epoxide.
There are three kinds of mycolates in the M. tuberculosis
cell wall, and they are R-mycolates, methoxymycolates, and
ketomycolates (Figure 1). Both the R- and methoxymycolates
only have the cis-cyclopropyl group at the proximal position,
while 17% of the cyclopropyl groups at the keto proximal
position are trans.18 Fifty-one percent of the total mycolates
are R-mycolates, 36% are methoxymycolates, and 13% are
ketomycolates.19 All three mycolates contain 24 and 26
carbon R branches in approximately the ratio of 10:90, with
negligible amounts of 22 carbon R branches. Methoxy-
mycolates and ketomycolates have longer mero chains than
R-mycolates. The total carbon numbers for the R, methoxy,
© 2004 American Chemical Society
Mycobacterium tuberculosis Cell Wall Permeability
Figure 1. Structures of mycolic acids found in the M. tuberculosis cell wall. The usual value for k is 23. The sum of n, m, and h is approximately
50. The shorter chain is the R branch, and the longer one is the mero chain. Both R and β carbons have R chirality. (See text for details.)
Reprinted with permission from Hong and Hopfinger (2004). Copyright American Chemical Society.
and keto forms are 76-82, 83-90, and 84-89, respec-
tively.20
AG, the other main component of the mycobacterial cell
wall, is a polysaccharide consisting of arabinose (Ara) and
galactose (Gal).21-23 Within AG, all arabinose and galactose
residues are in the furanose form, and mycolic acids are
located in clusters of four on the terminal hexaarabinofura-
noside through 1,5 linkages. However, only two-thirds of
the terminal arabinose residues are mycolated. The linker
disaccharidephosphate connects the galactan region of AG
to a peptidoglycan. Figure 2 provides an overview of the
structure and organization of the mycolyl-AG peptidoglycan
complex.
In the physical organization model for the mycobacterial
cell wall proposed by Minnikin, the mycobacterial cell wall
is composed of an asymmetric lipid bilayer (Figure 3).11 The
inner leaflet contains mycolic acids covalently linked to AG.
In this leaflet, mycolic acid hydrocarbon chains are tightly
packed in a parallel fashion, and predominantly oriented in
a direction perpendicular to the cell wall surface. The outer
leaflet contains the extractable lipids that accommodate the
uneven surface of the inner leaflet caused by the two
branches of mycolic acid that are not equal in length. It has
been shown that the outer leaflet is moderately fluid, while
the inner leaflet has very low fluidity.24 This low fluidity of
the inner leaflet may be the result of the unique chemical
structures of mycolic acids and their tight packing, which,
in turn, could account for the low permeability of the overall
cell wall. The cell wall permeability barrier prevents drugs
Biomacromolecules, Vol. 5, No. 3, 2004 1067
from entering the cell, which attributes, at least in part, to
the intrinsic drug resistance of mycobacteria.7
In the study reported in the preceding paper, conforma-
tional search and molecular dynamics simulation (MDS)
were employed to investigate the conformational profile of
AG and subsequently the conformations and structural
organization of the mycolyl-AG complex.25 An inner leaflet
molecular model of the M. tuberculosis cell wall was
constructed. (See the summary given in the Methods section.)
The mycolate hydrocarbon chains were determined to be
tightly packed and perpendicular to the “plane” formed by
the oxygen atoms of the 5-hydroxyl groups of the terminal
arabinose residues to which the mycolic acids bind. The
average packing distance between mycolic acids of M.
tuberculosis was estimated to be approximately 7.3 Å. This
previous computational study provides support for the
Minnikin model and also presents a way to probe the
mechanism of low permeability of the cell wall and, in turn,
the intrinsic drug resistance of M. tuberculosis.
To investigate M. tuberculosis cell wall permeability to
organic solutes in a quantitive manner and to gain more
insight into the nature of this barrier, a variety of drugs were
modeled as solutes undergoing transport across the M.
tuberculosis cell wall. The properties of the solute-
membrane complexes were studied by the means of MDS,
especially the diffusion coefficients of the solute molecules
inside the cell wall. For the purpose of comparing the
permeability of the M. tuberculosis cell wall to those of
bacterial and animal membranes, models for the complexes
1068 Biomacromolecules, Vol. 5, No. 3, 2004
Figure 2. Overview of the mycolyl-AG peptidoglycan complex (with permission from the Annual Review of Biochemistry, Volume 64, copyright
1995 by Annual Reviews www.annualreviews.org; minor modifications made).
Figure 3. Mycobacterial cell wall model proposed by Minnikin. (The
picture is taken from www.niaid.nih.gov/newsroom/focuson/tb02/
target.htm with modifications.) The funnel-shape structure in the
center of the cell wall represents a porin (for the description of the
function of porins, see ref 25).
of each of the same set of solutes and monolayers of
dipalmitoylphosphatidylethanolamine (DPPE) and dimyris-
toylphosphatidylcholine (DMPC) were also constructed and
studied, respectively. In addition, molecular similarity mea-
sures among the solutes were computed with the hope of
learning more about the solute features that govern perme-
ability across the M. tuberculosis cell wall.
Methods
1. Construction of a M. tuberculosis Cell Wall Model.
The procedures in the construction of a M. tuberculosis cell
wall model have been reported in the preceding paper and
are only summarized here to facilitate understanding of this
Hong and Hopfinger
study.25 As mentioned in the Introduction, the innermost part
of the cell wall, consisting solely of mycolic acids, is believed
to be the most tightly packed membrane region and has the
lowest permeability of the cell wall. Therefore, the mycolic
acid region was the focus in the construction of the M.
tuberculosis cell wall. Pseudo-mycolic acids (PMAs) were
designed to model this cell wall region (Figure 4). There
are 24 carbons in the R branch of a PMA molecule. The
mero chain was shortened in the modeling to the same length
as the R branch. A cis-cyclopropyl group was chosen to be
the functional group at the proximal position. The advantages
of this structural design are (1) the portions of the mycolic
acids that are located in the innermost cell wall, and
presumably are responsible for the cell wall low permeability,
are adequately represented by PMA molecules and (2) such
a PMA molecule represents all of the major types of mycolic
acids present in M. tuberculosis because these molecules have
similar structures to one another except for the distal moieties
(Figure 1) and a cis-cyclopropyl group is the dominant
functional group at the proximal position.
Even though it is known that the mycolic acids are very
tightly packed in the inner leaflet, there is very little data
available regarding the actual packing (including distance)
between mycolic acids. Therefore, the AG complex, the
polysaccharides to which the mycolic acids form ester bonds,
was studied. Conformational analysis was applied. However,
one AG complex is composed of three arabinans and one
galactan, and nearly 100 sugar residues are involved.
Consequently, there are hundreds of torsion angle degrees
of conformational freedom. A complete systematic torsion
angle conformational search of such a complex is not
realistic. Thus, the branch containing the 17 arabinose
Mycobacterium tuberculosis Cell Wall Permeability
Figure 4. Chemical structure of a PMA molecule. Carbon atoms located at position 13 on both the R and the mero chains are pointed out.
residues that are close to the termini of the arabinan chain
and covalently linked to the mycolic acids was chosen to
represent the structural features of the AG complex (Figure
2). Exhaustive random conformational sampling was per-
formed on this structure. The conformations generated were
then ranked according to their conformational potential
energies. The low energy conformations were chosen to form
complexes with the PMAs. Energy minimization was per-
formed on the resultant PMA-AG complexes, and the
complex with the lowest conformational potential energy was
used as a starting point in extensive MDS. The average
distance between the oxygen atoms of the 5-OH groups of
the terminal arabinose residues, where the mycolic acids bind,
was calculated on the basis of the MD trajectories taken after
the complex had reached equilibrium. Because the mycolic
acids exist as mycolates in the inner leaflet, a pseudo-
mycolate (PM) was generated by esterifying the carboxyl
group of a PMA molecule with methanol. A PM monolayer
was then constructed using the average packing distance
between the oxygens of the 5-OH groups that, in turn,
represents the inner leaflet of the M. tuberculosis cell wall.
2. Construction of the 13 Solutes. Thirteen known drugs
were selected as solutes to study the permeability of the M.
tuberculosis cell wall. These drugs include seven first- and
second-line anti-TB drugs, five general antimycobacterial
drugs, and one compound that has no antimycobacterial
activity. The structures of these 13 “solutes” are given in
Figure 5. The HyperChem 6.01 software26 was used to build
the three-dimensional structures of the solutes. The MM+
molecular mechanics force field, also implemented in Hy-
perChem, was utilized in the solute structure optimizations.
Each structure was subject to conjugate gradient energy
minimization using the Polak-Ribiere first derivative
method,27 until a derivative convergence criterion of 0.1 kcal/
(mol Å) was reached or until a maximum number of
iterations were performed. The HyperChem software sets the
maximum number of iterations to be 15 times the number
of atoms in the system. After each structure was energy
minimized in the MM+ force field, it was further energy
minimized by the AM1 semiempirical method.28 Partial
atomic charges were estimated by performing a single point
AM1 calculation on the minimized structure.
3. Construction and MDS of the Complexes of the M.
tuberculosis Cell Wall Model with the 13 Solutes. On the
basis of the estimated average packing distance between
mycolic acids from our previous study, the unit cell
parameters of the PM monlayer were selected to be a ) 7.3
Å, b ) 7.3 Å, and γ ) 90°.25 Each solute was inserted into
the M. tuberculosis cell wall, modeled by the PM monolayer,
as is illustrated in Figure 6. The solutes were located in the
region defined by the second and third rows and the third
Biomacromolecules, Vol. 5, No. 3, 2004 1069
and fourth columns of the PMs. The primary (largest) inertial
axis of each solute was aligned parallel to the Z axis of the
mycolates (i.e., the direction of the mycolate hydrocarbon
chains). The solute was then moved along the Z axis until
its center of mass was in the “plane” formed by the carbon-
13 atoms of the mycolates’ mero chains; see Figure 4.
The 13 cell wall-solute complexes generated were each
subjected to extensive MDS. The MOLSIM package,29 with
an extended MM2 force field,30,31 was used to perform the
MDS. The simulation temperature was set at 311 K and held
constant during the MDS by coupling the system to an
external fixed temperature bath.32 The initial periodic bound-
ary conditions were set at 36.5 Å for the X and Y dimensions,
80 Å for the Z dimension, and 90° for the surface orientation
angle, γ. The simulation sampling time was 140 ps with
intervals of 0.001 ps for a total sampling of 140 000
conformations for each of the complexes. Geometries gener-
ated during the MDS were recorded every 0.1 ps. Three
different random seeds were used to initiate the MDS. The
MDS routine of the MOLSIM software package, as described
above, was applied throughout this study unless stated
otherwise. Methyl ester carbons and the terminal carbons of
the PM mero chains were each assigned heavy masses of
1000 amu. The purpose of the heavy mass assignment is to
take the influence of additional bonded structure into
consideration. For example, restrictions in the movement of
the terminal carbons would occur in a MDS if the mero
chains had not been artificially shortened.
The displacement, ∆di, between two adjacent time steps
was determined for each atom of each solute from its MD
trajectory, and the individual diffusion coefficient of every
solute atom was calculated using the mean-square displace-
ment method.33 The mean-square displacement is given as
n
∑
i)1
∆di2
χ2 ) (1)
n
where n is the total number of time steps in MDS. From the
Einstein diffusion equation, the diffusion coefficient, D, is
given by
χ2
D) (2)
2∆t
where ∆t is the size of one time step.34 The diffusion
coefficient of the complete solute molecule is then calculated
as the average of the diffusion coefficients of all atoms of
the solute.
4. Construction and MDS of the Complexes of the
DPPE Monolayer Model with the Solutes. A DPPE
1070 Biomacromolecules, Vol. 5, No. 3, 2004
Figure 5. Chemical structures of the 13 solute molecules considered in this study.
monolayer model was previously built25 (see Figure 7A for
the structure of DPPE). The complexes of the DPPE
monolayer with the thirteen solutes were constructed using
the MI-QSAR software.35 To prevent unfavorable van der
Waals interaction between the solute molecule and the
membrane DPPE, one of the “center” DPPE molecules was
removed from the monolayer model and the solute molecule
was inserted in the space created by the missing DPPE
molecule. This procedure was not applied in the construction
of the solute-PM complex model. In the previous study,25
it was found that the average packing distance of mycolic
acids in a PM monolayer is correspondingly less than that
of DPPEs in a DPPE monolayer, and the inner leaflet of the
Hong and Hopfinger
M. tuberculosis cell wall exists in a relatively highly ordered
state. Hence, the solutes were inserted into an intact PM
monolayer model to mimic this dense and highly ordered
state.
Each solute molecule was placed between the regions of
head groups and aliphatic chains of the monolayer with the
most polar group of the solute molecule “facing” toward the
head group region of the monolayer. MDSs were performed
on the monolayer-solute complex using MOLSIM. The
periodic boundary conditions, which were used to build the
DPPE monolayer model, were employed (a ) 37.5 Å, b )
37.5 Å, c ) 80 Å, and γ ) 92°).25 The monolayer-solute
complexes were simulated for 70 ps using time intervals of
Mycobacterium tuberculosis Cell Wall Permeability
Figure 6. Illustration of an initial insertion, position, and alignment
of a solute molecule into the M. tuberculosis cell wall model using
ethambutol as an example.
0.001 ps. The diffusion coefficients of the solutes in the
DPPE model membrane were then calculated in the same
manner as just described for the M. tuberculosis cell wall
model.
5. Construction and MDS of the Complexes of the
DMPC Monolayer Model with the Solutes. The complexes
of the DMPC monolayer model with the solute molecules
were constructed in the same manner as in step 4 for DPPE
(see Figure 7B for the structure of DMPC). The MDSs were
carried out using the MOLSIM package. The periodic
boundary conditions employed are a ) 40 Å, b ) 40 Å, c
) 80 Å, and γ ) 96°. The simulation sampling time was 70
ps using time intervals of 0.001 ps. Diffusion coefficients
Figure 7. Chemical structures of DPPE and DMPC.
Biomacromolecules, Vol. 5, No. 3, 2004 1071
Table 1. IPEs Used in the Molecular Similarity Analyses
IPE code symbol definitions
0 ALL all atoms in the molecule
1 NP nonpolar atoms
2 P+ polar atoms with positive charge
3 P- polar atoms with negative charge
4 HBA hydrogen bond acceptor atoms
5 HBD hydrogen bond donor atoms
6 ARO aromatic atoms
7 HS non-hydrogen atoms (hydrogen suppressed)
were calculated for the solute molecules in the same manner
as in PM and DPPE.
Additional information regarding the construction of a
phospholipid monolayer, and its complexes with solutes,
using the MI-QSAR software can be found in refs 36-38.
6. Molecular Similarity Analysis of the 13 Solutes. The
4D-MS module of the 4D-QSAR software package was used
to perform the four-dimensional molecular similarity study
of the 13 solutes.39,40 4D-QSAR molecular similarity analysis
considers both the three-dimensional molecular structure and
the corresponding complete conformational ensemble of
states of a molecule in estimating molecular similarity. The
formalism measures relative molecular similarity (RMS) that
is dependent upon an alignment constraint (an external
reference frame) and absolute molecular similarity (AMS)
that is alignment-independent. This method also allows
estimation of molecular similarity measures on the basis of
atom types composing molecules. 4D-QSAR defines eight
atom types and names them as interaction pharmacophore
elements, IPEs (Table 1). In this study, AMS analysis based
on the eight IPE types was applied. MDS was performed to
generate a conformational ensemble profile (CEP) for every
solute molecule. The MOLSIM package was again used to
perform the simulations. The simulation sampling time was
70 ps using time intervals of 0.001 ps. The atomic coordi-
nates of all conformations sampled for every solute were
recorded every 0.1 ps and stored in a trajectory file that, in
turn, constituted the corresponding CEP of the solute. For
the detailed methodology of 4D-MS formalism, see ref 40.
1072 Biomacromolecules, Vol. 5, No. 3, 2004 Hong and Hopfinger

Table 2. MW and Calculated log P (calcd log P) Values of the

Solute Molecules
solute MW calcd log Pa
First- and Second-Line Anti-TB Drugs
isoniazid 137.14 0.49
PAS 153.14 0.68
ethionamide 166.24 1.97
ethambutol 204.31 0.29
ciprofloxacin 331.34 0.67
ofloxacin 361.37 0.62
clofazimine 473.40 6.98
General Antimycobacterial Drugs
amithiazone 236.29 1.24
dapsone 248.30 1.31
hydnocarpic acid 252.40 4.78
chaulmoogric acid 280.45 5.57
thiocarlide 400.58 6.72
Non-Antimycobacterial Drug
phencycline 243.39 3.98
a Calcd log P values were calculated from the calcd log P module in
the HyperChem 6.01 package.

Figure 8. Three-dimensional structures of clofazimine, ethionamide,

and ethambutol after structure optimization: H (white), C (grey), N
(blue), Cl (green), S (yellow), and O (red).

Results

1. Permeation of the Solute Molecules through the M.

tuberculosis Cell Wall Model. The 13 solute molecules
considered in this study can be classified into three categories
according to their molecular shapes: (1) bulky, as represented
by clofazimine, and including clofazimine, phencycline,
ciprofloxacin, and ofloxacin, where the first two molecules
are much bulkier than the last two (phencycline, the only
non-antimycobacterial drug in the dataset, was chosen for
its bulky and “football”-like molecular shape.); (2) flat, as
represented by ethionamide, and including isoniazide, ethion-
amide, p-aminosalicyclic acid (PAS), amithiazone, dapsone,
and thiocarlide; and (3) linear, as represented by ethambutol,
and including ethambutol, hydnocarpic acid, and chaulmoo-
gric acid. The three-dimensional structures of the three
representative molecules, clofazimine, ethionamide, and
ethambutol, after structure optimization, are shown in Figure
8.
In addition to steric features, the molecular size, reflected
by molecular weight (MW), of the solute molecules also
varies significantly (Table 2). Isoniazid is the smallest
molecule in the dataset and has a MW of 137 amu, while
clofazimine has the highest MW of 473 amu. The solute
molecules also possess a considerable range in lipophilicity
(Table 2) and can be grouped into two classes based on their
log P values being greater (class 1) or less (class 2) than 2.
Clofazimine, hydnocarpic acid, chaulmoogric acid, thiocar-
lide, and phencycline belong to the first class, and the rest
of the compounds fall into the second class. Overall, this
Figure 9. Relationship between mean-square displacement of the
solute (10-20 cm2) and simulation time (ps). The thick green-colored
curve is made of data points. The regression line (in red) and fit are
shown. The diffusion coefficient calculated here corresponds to Dma
(2) in Table 3.
dataset of 13 solutes covers a relatively large and diverse
chemical space regarding molecular shape, size, and polarity.
Figure 9 illustrates the relationship between mean-square
displacements (χ2) of the solutes in the M. tuberculosis cell
wall model and simulation time (t), using ethambutol as an
example. The diffusion coefficients of the solute molecules,
calculated as 1/2 of the slope of the linear regression line of
χ2 versus time (see eq 2), are given in Table 3 in the form
of Dma, where “D” stands for diffusion coefficient and “ma”
stands for mycolic acid. Bulky solutes have a lower Dma than
flat and linear solutes, while there is no significant difference
between solutes in the latter two classes. Moreover, it is
difficult to find a straightforward relationship between Dma
values of the solutes and their molecular shapes and
polarities. The geometry of a M. tuberculosis cell wall-solute
complex, at the end of MDS, is exemplified using ethambutol
(Figure 10). The hydrocarbon chains of the mycolates remain
tightly packed and perpendicular to the cell surface through-
out the MDS.
Mycobacterium tuberculosis Cell Wall Permeability Biomacromolecules, Vol. 5, No. 3, 2004 1073

Table 3. Diffusion Coefficients (10-8 cm2/s) of the Solute

Molecules in the M. tuberculosis Cell Wall (Dma)a
solute 〈Dma〉 Dma (1) Dma (2) Dma (3)
isoniazid 1.0 1.1 0.9 1.0
PAS 1.0 1.0 1.0 1.0
ethionamide 1.0 0.9 1.0 1.0
ethambutol 1.3 1.2 1.3 1.3
ciprofloxacin 0.8 0.8 0.7 0.8
ofloxacin 0.7 0.7 0.7 0.7
clofazimine 0.7 0.8 0.6 0.7
amithiazone 0.9 0.8 1.0 0.9
dapsone 1.3 1.6 1.1 1.1
hydnocarpic acid 1.1 1.1 1.0 1.2
chaulmoogric acid 1.1 1.1 1.1 1.1
thiocarlide 1.1 1.2 1.1 1.1
phencycline 0.8 0.7 0.7 0.9
a The D
ma (1-3) were calculated from the molecular dynamics trajec-
tories using three different random seeds to inititate the MDS. The 〈Dma〉
values reported in the second column are the average values from the
three MDSs.

Figure 11. Axial ∆d2 (10-20 cm2) components of clofazimine,

Figure 10. Structure of the membrane-solute complex of the M.
tuberculosis cell wall with ethambutol at the end of MDS. The
hydrogen atoms of the cell wall are shown in white, the carbon atoms
in gray, and the oxygen atoms in red. The solute, ethambutol, is
shown in CPK form.
Some solutes during the MDS diffusion experiments
predominantly move inside the membrane in directions other
than through the cell. These molecules may have high
diffusion coefficients, but their actual uptake into the interior
of a cell is low. The measurement of a diffusion coefficient
cannot detect this lateral component to net diffusion. Thus,
the observed diffusion coefficient can, perhaps, be misleading
of the cellular uptake of the compound.
For the solutes in Figure 5, the individual movements
along the X, Y, and Z axes of each solute were calculated
from the distance that the molecule moves parallel to each
axis during every single MDS time step. These distances
ethambutol, and chaulmoogric acid in the membrane models over
the course of the MDS. The unit of simulation time is ps. The color
coding for the X, Y (lateral diffusion), and Z (transverse diffusion) is
given in the clofazimine plot.
are referred to as ∆dX2, ∆dY2, and ∆dZ2, respectively.
According to the different patterns of movement exhibited
by the solutes inside the cell wall model, the solute molecules
can be classified into three groups. Solutes in the first group
move in small ranges relative to each of the three axes. These
solutes are clofazimine, ciprofloxacin, ofloxacin, and phen-
cycline. The second group of solutes also moves uniformly
along all three axes, but the overall movement is larger than
that of solutes in the first group. Ethambutol, dapsone,
isoniazid, ethionamide, amithiazone, PAS, and thiocarlide
exhibit group 2 behavior. Chaulmoogric acid and hydno-
carpic acid have considerably different movements from the
solutes of groups 1 and 2. These two solutes do not move
equally along each axis, instead they undergo much more
movement in the membrane, the X and Y directions, than
through the membrane, the Z direction. Chaulmoogric acid
and hydnocarpic acid, thus, make up the third group of
solutes.
Clofazimine, ethambutol, and chaulmoogric acid are
representative of the three solute groupings with respect to
1074 Biomacromolecules, Vol. 5, No. 3, 2004 Hong and Hopfinger

Table 4. Components of the Diffusion Coefficients with Respect to the Individual Axes (10-8 cm2/s) for the Solute Molecules in the M.
tuberculosis Cell Walla
〈Dma〉 Dma (1) Dma (2) Dma (3)
solute 〈Dx〉 〈Dy〉 〈Dz〉 Dx Dy Dz Dx Dy Dz Dx Dy Dz
isoniazid 0.4 0.4 0.3 0.4 0.4 0.3 0.3 0.4 0.3 0.4 0.4 0.3
PAS 0.4 0.4 0.3 0.4 0.4 0.4 0.4 0.4 0.3 0.4 0.4 0.3
ethionamide 0.3 0.4 0.3 0.4 0.3 0.3 0.3 0.4 0.3 0.3 0.4 0.3
ethambutol 0.5 0.5 0.3 0.5 0.5 0.3 0.5 0.4 0.4 0.5 0.5 0.4
ciprofloxacin 0.3 0.3 0.2 0.3 0.3 0.3 0.3 0.2 0.2 0.3 0.3 0.2
ofloxacin 0.3 0.3 0.2 0.3 0.3 0.2 0.3 0.3 0.2 0.2 0.3 0.2
clofazimine 0.3 0.3 0.2 0.3 0.3 0.3 0.2 0.2 0.2 0.3 0.3 0.2
amithiazone 0.3 0.4 0.3 0.3 0.3 0.2 0.3 0.4 0.3 0.3 0.4 0.3
dapsone 0.5 0.5 0.4 0.6 0.5 0.6 0.4 0.4 0.4 0.4 0.4 0.3
hydnocarpic acid 0.4 0.5 0.2 0.4 0.5 0.2 0.4 0.5 0.2 0.5 0.5 0.2
chaulmoogric acid 0.5 0.5 0.2 0.5 0.4 0.2 0.4 0.6 0.2 0.4 0.5 0.2
thiocarlide 0.3 0.2 0.2 0.3 0.3 0.2 0.3 0.2 0.2 0.2 0.2 0.2
phencycline 0.3 0.3 0.2 0.3 0.3 0.2 0.3 0.3 0.2 0.2 0.3 0.2
a The indices, (1-3), represent the MDS experiments initiated by three different random seeds. The diffusion coefficients reported in the second to

fourth columns are the average values from the three MDS experiments. The diffusion coefficients of hydnocarpic acid and chaulmoogric acid, which
exhibit significant lateral diffusion, are shown in bold.

Table 5. Two-Dimensional and Four-Dimensional AMS Molecular

diffusion and their movement patterns are shown in Figure Similarity Measurements between Pairs of Solute Moleculesa
11. These movement patterns, especially those of the third
Part A. Two-Dimensional Molecular Similarity
group, can also be seen in Table 4, where Dma is partitioned chaulmoogric acid
into Dma-x, Dma-y and Dma-z. The Dma-z of hydnocarpic acid hydnocarpic acid
and chaulmoogric acid (in bold) are significantly smaller than 2D 0.906
the corresponding Dma-x and Dma-y (in bold). There is a
Part B. Four-Dimensional AMS
correlation between the steric shape of a molecule and its amithiazone isoniaid ciprofloxacin
movement pattern inside the M. tuberculosis cell wall. All dapsone PAS ofloxacin
the spherical and bulky molecules are in the first group, while ALL 0.938 0.935 0.901
linear and flat molecules compose the second and third isoniaid
groups. For every solute in the first and second groups, the PAS
extent of its movement relative to each reference axis is NP 0.961
proportional to its Dma value. However, the large Dma values chaulmoogric acid isoniazid isoniazid PAS
of the third group of solutes do not reflect their penetration hydnocarpic acid ofloxacin ethionamide amithiazone
rates. Chaulmoogric acid and hydnocarpic acid both resemble P+ 1.000 0.942 0.939 0.917
the shape, charge distribution, and polar/nonpolar asymmetry PAS isoniazid
of a PM molecule, which may promote their preferred lateral ethambutol ethambutol
movements within, as opposed to through, the membrane. P- 0.957 0.907
In fact, these two molecules are readily taken up into the clofazimine PAS PAS isoniazid
PM monolayer, and by the end of MDS the polar head groups ethambutol ethambutol clofazimine ethambutol
of these acid solutes are located in the same region as the HBA 0.961 0.957 0.937 0.907
head groups of mycolates. Lateral movement is, in fact, quite chaulmoogric acid ciprofloxacin amithiazone PAS
common for molecules composing membranes and includes hydnocarpic acid thiocarlide ethambutol amithiazone
phospholipids and membrane proteins.41,42 The first observa- HBD 1.000 0.954 0.925 0.917
tion of this event using MDS has been reported by Moore ofloxacin phencycline ciprofloxacin ciprofloxacin
and co-workers recently.43 PAS PAS ofloxacin phencycline
Table 5 lists the two-dimensional and four-dimensional ARO 1.000 0.999 0.999 0.999
(alignment independent) AMSs for the solute molecules. The ofloxacin isoniazid amithiazone amithiazone
AMS consists of measurements based on the eight IPE types. phencycline ethionamide PAS ciprofloxacin
Only molecule pairs having their similarity measures greater ARO 0.999 0.999 0.999 0.999
than 0.900 are considered similar in this study and are given amithiazone amithiazone PAS
in Table 5. Most spherical and bulky compounds are similar ofloxacin phencycline ciprofloxacin
to one another, such as ciprofloxacin and ofloxacin, having ARO 0.999 0.999 0.999
the ALL-AMS, ARO-AMS and HS-AMS measures PAS ciprofloxacin
greater than 0.900. Some flat-shaped molecules are also ethionamide ofloxacin
similar, including isoniazid and PAS, and amithiazone and HS 0.941 0.922
dapsone, and so forth. Chaulmoogric acid and hydnocarpic a A value of 0.000 means there is no similarity and a value of 1.000

acid are similar, having a two-dimensional molecular simi- implies identical similarity.
larity measure of 0.906 and the AMS P+-AMS and HBD- It is a core working hypothesis in medicinal chemistry that
AMS have measured similarities of 1.000. structurally similar molecules are likely to express similar
Mycobacterium tuberculosis Cell Wall Permeability Biomacromolecules, Vol. 5, No. 3, 2004 1075

Table 6. Diffusion Coefficients (10-8 cm2/s) of the Solute Molecules in the DPPE and DMPC Monolayers (DDPPE, DDMPC)
solute 〈DDPPE〉 DDPPE (1) DDPPE (2) DDPPE (3) 〈DDMPC〉 DDMPC (1) DDMPC (2) DDMPC (3)
isoniazid 1.5 1.6 1.5 1.5 1.3 1.4 1.2 1.3
PAS 1.6 1.7 1.5 1.5 1.2 1.3 1.3 1.1
ethionamide 1.5 1.6 1.5 1.5 1.3 1.3 1.3 1.4
ethambutol 1.7 1.8 1.6 1.6 1.6 1.6 1.6 1.7
ciprofloxacin 1.2 1.2 1.2 1.1 1.0 1.0 1.0 1.0
ofloxacin 1.0 1.0 1.1 0.9 0.9 0.9 0.9 1.0
clofazimine 1.0 1.0 0.9 1.0 0.8 0.8 0.9 0.8
amithiazone 1.7 1.8 1.5 1.7 1.4 1.3 1.4 1.3
dapsone 1.7 1.7 1.8 1.7 1.4 1.5 1.4 1.4
hydnocarpic acid 1.6 1.5 1.6 1.6 1.4 1.4 1.4 1.3
chaulmoogric acid 1.4 1.4 1.4 1.5 1.5 1.4 1.5 1.5
thiocarlide 1.3 1.3 1.4 1.3 1.2 1.1 1.2 1.2
phencycline 1.2 1.2 1.2 1.2 0.9 0.9 0.9 1.0

biological behavior profiles including absorption distribution,
metabolism, elimination, and toxicity (ADMET) properties.
In this case, it would appear that similar compounds have
similar permeation behavior within a common cell wall
environment. This conclusion is supported by several ex-
amples, such as chaulmoogric acid and hydnocarpic acid,
isoniazid and PAS, amithiazone and dapsone, and the
spherical and bulky solutes.
2. Comparison of Diffusion Coefficients from the DPPE
and the DMPC Monolayer Models. There is no significant
difference between the DDPPE and the DDMPC values for the
solute molecules studied (Table 6). But, as expected, solutes
in both phospholipid membrane models have correspondingly
larger diffusion coefficients than the Dma of the M. tuber-
culosis cell wall model. Most solutes retain three-dimensional
movement patterns in both the DPPE and DMPC monolayers
such as those observed in the M. tuberculosis cell wall model,
except for chaulmoogric acid. This molecule moves unifor-
mally with respect to the three reference axes in the DPPE
and DMPC monolayers. Only hydnocarpic acid from the
solute dataset retains its principal movement within, as
opposed to through, the membrane in the DPPE and DMPC
monolayers. That is, hydnocarpic acid undergoes significant
lateral diffusion in both DPPE and DMPC membranes.
Discussion
The molecular shape of a solute is found to be an important
factor for the permeation behavior of a solute through the
M. tuberculosis cell wall. For example, clofazimine and
thiocarlide have similar MWs. However, bulky-shaped
clofazimine has a Dma value significantly lower than that of
the flat-shaped thiocarlide. Moreover, phencycline and clo-
fazimine have the same Dma values even though the MW of
phencycline is only half that of clofazimine.
A major finding of this MDS study is the observation for
some solutes of predominant lateral diffusion within, as
opposed to transverse diffusion across, the M. tuberculosis
cell wall model. Chaulmoogric acid is one of the solutes that
prefers to move “parallel” to the cell surface inside the cell
wall structure (lateral diffusion) as opposed to moving along
the hydrocarbon chains of mycolates (transverse diffusion).
As a result of preferred lateral diffusion, such solutes remain
inside the cell wall. Overall, the apparent penetration rate
and the corresponding absorption of solute inside the cell
are both low even though the diffusion coefficient may be
large.
Lateral diffusion may also affect cell wall function. In fact,
lipid analysis of M. Vaccae grown in the presence of
chaulmoogric acid has demonstrated that chaulmoogric acid
is taken up by the organism and incorporated into cell wall
components.44 It is also observed that cell growth is retarded
by the addition of chaulmoogric acid to the growth medium.44
Therefore, it is interesting to postulate that the antimyco-
bacterial properties of chaulmoogric acid result from its
uptake into or distortion of the cell wall structure. In the
bacterial and animal membranes modeled by DPPE and
DMPC monolayers, respectively, chaulmoogric acid under-
goes transverse diffusion across the membrane equally well
compared to lateral diffusion. Thus, this compound is not
readily retained inside the phospholipid membranes and its
cellular absorption and distribution properties are likely to
reflect its diffusion coefficients in these membranes. Al-
though chaulmoogric acid is traditionally used to treat
leprosy, this molecule might also provide a promising anti-
TB mechanism: the uptake into or disruption of the M.
tuberculosis cell wall structure. Thus, chaulmoogric acid
could be an important lead for future anti-TB drug develop-
ment. Hydnocarpic acid, similar to chaulmoogric acid, is also
used to treat leprosy, and its antimycobacterial effects may
also be due to the uptake into or distortion of the mycobac-
terial cell wall structure.
The resolution of solute diffusion into transverse and lateral
components in membranes might be very important in
studying ADMET properties of drug candidates. It is shown
that lateral diffusion plays a significant role in the transport
of molecules across tissues, including the blood-brain
barrier, corneal membrane, and intestinal membranes (Figure
12).45 However, to get inside a target cell and exert its action,
the drug candidate has to have good transverse diffusion
properties. Among the solute molecules currently studied,
hydnocarpic acid is the only molecule that displays pre-
dominant lateral diffusion in the bacterial and animal
membrane models, while the rest of the solutes exhibit,
essentially, equal lateral and transverse diffusion. Considering
that these solute molecules are actual drugs and have
clinically acceptable ADMET properties, it is not clear from
the results of this study what the delicate balance between
1076 Biomacromolecules, Vol. 5, No. 3, 2004
Figure 12. Illustration of lateral diffusion and transverse diffusion
processes across three cells of a tissue. The green circles represent
the plasma membrane. The red arrows represent nonpolar solutes,
and the red circle stands for the lateral diffusion of the solute
molecules inside the membrane. The diffusion processes are driven
by the concentration gradient of the solute molecules.
lateral and transverse diffusion should be to realize an
effective drug for a particular therapy.
The molecular similarity analysis suggests that compounds
having markedly high features of molecular similarity will
display similar transport behavior in a common membrane/
cell wall environment. What is not clear is if a set of 13
diverse compounds is sufficient to extend the similar
structure-similar transport behavior observation into a design
rule. Moreover, there remains the nagging problem that, in
general, building in favorable ADMET properties in a drug
candidate correspondingly diminishes biological potency.
Dma is generally smaller than both DDPPE and DDMPC,
indicating it is more difficult for a solute to diffuse through
a tightly packed M. tuberculosis cell wall inner leaflet than
to pass through a more liquid-crystal-like phospholipid
membrane structure. This observation is consistent with the
results from the M. tuberculosis cell wall construction study.25
However, the calculated variations in diffusion coefficients
of the same solute in the different membrane/cell wall
environments and the computed diffusion coefficients of the
different solutes in the same membrane/cell wall environment
are markedly small (1 order of magnitude) in terms of both
the different membrane environments and the diversity in
chemical structures of the solutes. The role AG plays in
formation and stabilization of the M. tuberculosis cell wall
may not have been adequately modeled and could be
responsible for this small range, overall, in the computed
diffusion coefficients (Figures 2 and 3).25 When a bulky
solute is inserted into the cell wall model, it introduces
significant deformations to the structure of the M. tubercu-
losis model cell wall. For example, in the cell wall-
clofazimine complex, the PM molecules in the center of the
cell wall are “squeezed” out by clofazimine, and as a result,
the “surface” formed by the head groups of the PM molecules
is no longer flat (Figure 13). The “chain reaction” induced
by this structural deformation might disrupt the extensive
inter-residue and intra-residue hydrogen bonds among the
hydroxyl groups of arabinose and galactose residues of AG.25
In turn, changes in this hydrogen bonding pattern might alter
the stability of the original mycolyl-AG complex structure.
Consequently, the mycolyl-AG complex likely tries to keep
the solute molecule out and “push” it back to the relatively
Hong and Hopfinger
Figure 13. Structure of the membrane-solute complex of the M.
tuberculosis cell wall with clofazimine at the end of MDS. Molecules
are represented in the same manner as in Figure 10.
fluid outer leaflet where the solute first enters the cell wall.
This “pushing back” likely involves some kind of collabora-
tive movement between the AG backbone and the mycolate
chains. As a result, the permeation rates of the solutes will
be reduced. However, in the MDS experiments conducted
in this study, the impact of cell wall structure deformation
on solute diffusion processes may not be adequately modeled.
Another possible reason for the small variation in com-
puted diffusion coefficients may arise from neglect of the
concentration gradients of the solute molecules in the MDS
experiments. That is, the solubility of the solute in the
membrane (a component of permeation) has not been
considered. However, the variations in the computed diffu-
sion coefficients, although small, are still consistent with the
observed trends in the biological behavior of the different
solute molecules in the different membrane/cell wall envi-
ronments. For example, the differences in the Dma values of
clofazimine, ethambutol, and chaulmoogric acid reveal quite
different behaviors among these solutes in the M. tuberculosis
cell wall model (Tables 3 and 4, Figure 11).
In future work, a model for the entire mycolyl-AG
complex will be constructed and used to simulate the
permeability of the M. tuberculosis cell wall. A means to
include the concentration gradient of the solutes in the
membrane/cell wall (solute solubility in the membrane)
should also be developed to actually represent permeation,
as opposed to diffusion, processes for the solutes.
Acknowledgment. Resources of the Laboratory of Mo-
lecular Modeling and Design were used in performing this
study. We gratefully appreciate financial support from The
Mycobacterium tuberculosis Cell Wall Permeability
Chem21 Group, Inc., and X.H. acknowledges a University
Fellowship from UIC. We also thank Mr. Jianzhong Liu of
our research group and Professor Scott G. Franzblau of the
Institute for Tuberculosis Research at UIC for their helpful
comments and discussions over the course of this work.
References and Notes
(1) WHO Tuberculosis Fact Sheet. http://www.who.int/mediacentre/
factsheets/who104/en/index.html (accessed May 2003).
(2) WHO report 2003. http://www.who.int/gtb/publications/globrep/
index.html (accessed June 2003).
(3) Dolin, P. J.; Raviglione, M. C.; Kochi, A. Global tuberculosis
incidence and mortality during 1990-2000. Bull. W. H. O. 1994,
72, 213-220.
(4) Jarlier, V.; Nikaido, H. Mycobacterial cell wall: structure and role
in natural resistance to antibiotics. FEMS Microbiol. Lett. 1994, 123,
11-18.
(5) Brennan, P. J.; Nikaido, H. The envelope of mycobacteria. Annu.
ReV. Biochem. 1995, 64, 29-63.
(6) Barry, C. E., III; Mdluli, K. Drug sensitivity and environmental
adaptation of mycobacterial cell wall components. Trends Microbiol.
1996, 4, 275-281.
(7) Jarlier, V.; Nikaido, H. Permeability barrier to hydrophilic solutes
in Mycobacterium chelonei. J. Bacteriol. 1990, 172, 1418-1423.
(8) Trias, J.; Benz, R. Permeability of the cell wall of Mycobacterium
smegmatis. Mol. Microbiol. 1994, 14, 283-290.
(9) Chambers, H. F.; Moreau, D.; Yajko, D.; Miick, C.; Wagner, C.;
Hackbarth, C.; Kocagoz, S.; Rosenberg, E.; Hadley, W. K.; Nikaido,
H. Can penicillins and other beta-lactam antibiotics be used to treat
tuberculosis? Antimicrob. Agents Chemother. 1995, 39, 2620-2624.
(10) Jarlier, V.; Gutmann, L.; Nikaido, H. Interplay of cell wall barrier
and beta-lactamase activity determines high resistance to beta-lactam
antibiotics in Mycobacterium chelonae. Antimicrob. Agents Chemoth-
er. 1991, 35, 1937-1939.
(11) Minnikin, D. E. In The biology of the Mycobacteria; Ratledge, C.,
Standford, J. L., Eds.; Academic: London, 1982; Vol. 1, pp 95-
184.
(12) Minnikin, D. E.; Goodfellow, M. In Microbiological classification
and identification; Goodfellow, M., Board, R. G., Eds.; Academic:
London, 1980; pp 189-256.
(13) Dobson, G.; Minnikin, D. E.; Minnikin, S. M.; Parlett, J. H.;
Goodfellow, M. In Chemical methods in bacterial systematics;
Goodfellow, M., Minnikin, D. E., Eds.; Academic: London, 1985;
pp 237-265.
(14) Brennan, P. J. In Microbial lipids; Ratledge, C., Wilkinson, S. G.,
Eds.; Academic: London, 1988; Vol. 1, pp 203-298.
(15) Brennan, P. J. Structure of mycobacteria: recent developments in
defining cell wall carbohydrates and proteins. ReV. Infect. Dis. 1989,
11 (Suppl.), 420-430.
(16) Hunter, S. W.; Murphy, R. C.; Clay, K.; Goren, M. B.; Brennan, P.
J. Trehalose-containing lipooligosaccharides. A new class of species-
specific antigens from mycobacterium. J. Biol. Chem. 1983, 258,
10481-10487.
(17) Hunter, S. W.; Gaylord, H.; Brennan, P. J. Structure and antigenicity
of the phosphorylated lipopolysaccharide antigens from the leprosy
and tubercle bacilli. J. Biol. Chem. 1986, 261, 12345-12351.
(18) Liu, J.; Barry, C. E., III; Besra, G. S.; Nikaido, H. Mycolic acid
structure determines the fluidity of the mycobacterial cell wall. J.
Biol. Chem. 1996, 271 (47), 29545-29551.
(19) Yuan, Y.; Crane, D. C.; Musser, J. M.; Sreevatsan, S. Barry, C. E.,
III. MMAS-1, the branch point between cis- and trans-cyclopropane-
containing oxygenated mycolates in Mycobacterium tuberculosis. J.
Biol. Chem. 1997, 272 (15), 10041-10049.
(20) Barry, C. E., III; Lee, R. E.; Mdluli, K.; Sampson, A. E.; Schroeder,
B. G.; Slayden, R. A.; Yuan, Y. Mycolic acids: structure, biosynthesis
and physiological functions. Prog. Lipid Res. 1998, 37, 143-179.
(21) Daffe, M.; Brennan, P. J.; McNeil, M. Predominant structural features
of the cell wall arabinogalactan of Mycobacterium tuberculosis as
revealed through characterization of oligoglycosyl alditol fragments
by gas chromatography/mass spectrometry and by 1H and 13C NMR
analysis. J. Biol. Chem. 1990, 265, 6734-6743.
Biomacromolecules, Vol. 5, No. 3, 2004 1077
(22) McNeil, M.; Daffe, M.; Brennan, P. J. Evidence for the nature of
the link between the arabinogalactan and peptidoglycan of myco-
bacterial cell walls. J. Biol. Chem. 1991, 265, 18200-18206.
(23) McNeil, M.; Daffe, M.; Brennan, P. J. Location of the mycolyl ester
substituents in the cell walls of mycobacteria. J. Biol. Chem. 1991,
266, 13217-13223.
(24) Liu, J.; Rosenberg, E. Y.; Nikaido, H. Fluidity of the lipid domain
of cell wall from Mycobacterium chelonae. Proc. Natl. Acad. Sci.
U.S.A. 1995, 92, 11254-11258.
(25) Hong, X.; Hopfinger, A. J. Construction, Molecular Modeling, and
Simulation of Mycobacterium tuberculosis Cell Walls. Biomacro-
molecules, 2004, 5, 1052-1065.
(26) HyperChem Program Release 6.01 for Windows; Hypercube, Inc.:
Gainesville, FL, 2000.
(27) Polak, E. Computational Methods in Optimization; Academic Press:
New York, 1971.
(28) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J. J. P.
AM1: a new general purpose quantum mechanical model. J. Am.
Chem. Soc. 1985, 107, 3902-3909.
(29) Doherty, D. C. MOLSIM User’s Guide; The Chem21 Group, Inc.;
Lake Forest, IL, 1997.
(30) Allinger, N. L. Conformational analysis. 130. MM2. A hydrocarbon
force field utilizing V1 and V2 torsional terms. J. Am. Chem. Soc.
1977, 99, 8127-8134.
(31) Hopfinger, A. J.; Pearlstein, R. A. Molecular mechanics force-field
parametrization precedures. J. Comput. Chem. 1984, 5, 486-492.
(32) Berendsen, H. J. C.; Postman, J. P. M.; Gunsteren, W. F. V.; Nola,
A. D.; Haak, J. R. Molecular dynamics with coupling to an external
bath. J. Chem. Phys. 1984, 81, 3684-3690.
(33) Tinoco, I.; Sauer, K.; Wang, J. Physical Chemistry. Prentice-Hall:
New York, 1985.
(34) Einstein, A. InVestigations on the theory of the Brownian moVement;
Methuen & Co., Ltd.: London, 1926.
(35) MI-QSAR User’s Manual, version 2.0; The Chem21 Group, Inc.; Lake
Forest, IL, 1997.
(36) Kulkarni, A.; Hopfinger, A. J. Membrane-interaction QSAR analy-
sis: application to the estimation of eye irritation by organic
compounds. Pharm. Res. 1999, 16 (8), 1245-1253.
(37) Kulkarni, A.; Han, Y.; Hopfinger, A. J. Predicting caco-2 cell
permeation coefficients of organic molecules using membrane-
interaction QSAR analysis. J. Chem. Inf. Comput. Sci. 2002, 42, 331-
342.
(38) Iyer, M.; Mishru, R.; Han, Y.; Hopfinger, A. J. Predicting blood-
brain barrier partitioning of organic molecules using membrane-
interaction QSAR analysis. Pharm. Res. 2002, 19 (11), 1611-1621.
(39) Hopfinger, A. J.; Wang, S.; Tokarski, J. S.; Jin, B.; Albuquerque,
M.; Madhav, P. J.; Duraiswami, C. Construction of 3D-QSAR models
using the 4D-QSAR analysis formalism. J. Am. Chem. Soc. 1997,
119, 10509-10524.
(40) Duca, J. S.; Hopfinger, A. J. Estimation of molecular similarity based
on 4D-QSAR analysis: formalism and validation. J. Chem. Inf.
Comput. Sci. 2001, 41, 1367-1387.
(41) Elson, E. L. Membrane dynamics studied by fluorescence correlation
spectroscopy and photobleaching recovery. Soc. Gen. Physiol. Ser.
1986, 40, 367-383.
(42) Frye, C. D.; Edidin, M. The rapid intermixing of cell surface antigens
after formation of mouse-human heterokaryons. J. Cell Sci. 1970, 7,
319-335.
(43) Moore, P. B.; Lopez, C. F.; Klein, M. L. Dynamical properties of a
hydrated lipid bilayer from a multinanosecond molecular dynamics
simulation. Biophys. J. 2001, 81, 2484-2494.
(44) Cabot, M. C.; Goucher, C. R. Chaulmoogric acid: assimilation into
the complex lipids of mycobacteria. Lipids 1981, 16 (2), 146-148.
(45) Edwards, A.; Prausnitz, M. R. Predicted permeability of the cornea
to topical drugs. Pharm. Res. 2001, 18 (11), 1497-1508.
BM0345155