NeuroRx威: The Journal of the American Society for Experimental NeuroTherapeutics

Medicinal Chemical Properties of Successful Central Nervous

System Drugs

Hassan Pajouhesh* and George R. Lenz†

*Department of Medicinal Chemistry, Neuromed Technologies Inc., Vancouver, British Columbia, Canada V6T 1Z4; and
†
GRLEN R&D Associates, Andover, Massachusetts 01810-5402

Summary: Fundamental physiochemical features of CNS to be optimized simultaneously with potency, selectivity, and
drugs are related to their ability to penetrate the blood-brain other biological parameters. The balance between optimizing
barrier affinity and exhibit CNS activity. Factors relevant to the the physiochemical and pharmacokinetic properties to make the
success of CNS drugs are reviewed. CNS drugs show values of best compromises in properties is critical for designing new
molecular weight, lipophilicity, and hydrogen bond donor and drugs likely to penetrate the blood brain barrier and affect
acceptor that in general have a smaller range than general relevant biological systems. This review is intended as a guide
therapeutics. Pharmacokinetic properties can be manipulated to designing CNS therapeutic agents with better drug-like prop-
by the medicinal chemist to a significant extent. The solubility, erties. Key Words: CNS drug, ADME (absorption, distribu-
permeability, metabolic stability, protein binding, and human tion, metabolism, and excretion), physicochemical properties,
ether-ago-go-related gene inhibition of CNS compounds need pharmacokinetic properties, blood-brain barrier (BBB), QSAR.

INTRODUCTION of inadequate pharmacokinetics of both developmental

and marketed drugs is failure in advanced development
The blood-brain barrier (BBB) and its penetration by
and/or market withdrawal.3
neurotherapeutics was the subject of a recent issue of this
Before the introduction of high-throughput screening
journal.1 This series of articles primarily had a biological
orientation. From a medicinal chemical perspective, the (HTS) and combinatorial screening libraries, most
ability to design drugs capable of penetrating the BBB screening for new CNS drugs was performed using bio-
and effecting the desired biological response is a formi- chemical methods, followed quickly by descriptive ani-
dable challenge. On the other hand, peripherally acting mal studies that could indicate whether or not a com-
drugs need to possess specific physical-chemical proper- pound penetrated the BBB. A lead compound would then
ties that prevent them from crossing the BBB. be profiled against receptor panels to look for selectivity.
In the broadest sense, moderately lipophilic drugs For instance, a synthetic derivative ABT-594, a very
cross the BBB by passive diffusion and the hydrogen potent nonopiate analgesic,4 derived from the Central
bonding properties of drugs can significantly influence American tree frog compound epibatidine,5 was profiled
their CNS uptake profiles. Polar molecules are generally against more than 70 receptors and ion channels before
poor CNS agents unless they undergo active transport clinical investigation. With the introduction of HTS and
across the CNS. Size, ionization properties, and molec- subsequent huge combinatorial libraries to feed it, whole
ular flexibility are other factors observed to influence animal testing has been even further removed from the
transport of an organic compound across the BBB.2 Ad- initial discovery process. Adding in informatics, genom-
ditionally, an effective drug possesses an appropriate ics, and proteomics to the discovery process makes ani-
pharmacokinetic profile. One of the major consequences mal testing even more remote from the initial discovery
process. In fact, significant resources can be invested in
a lead compound and series before even whole cell as-
says are conducted.6
Address correspondence and reprint requests to Hassan Pajouhesh, Despite the significant changes in drug discovery,
Ph.D., Neuromed Technologies Inc., 301-2389 Health Sciences Mall,
Vancouver, British Columbia, Canada V6T 1Z4. E-mail: hpajouhesh@ multiple classes of compounds affecting CNS processes
neuromed.com. at various intervention points are successfully used clin-

Vol. 2, 541–553, October 2005 © The American Society for Experimental NeuroTherapeutics, Inc. 541
542 PAJOUHESH AND LENZ

ically, and many more are in development.7 These com- On a molecular level, the BBB is not homogenous but
pounds possess tremendous chemical diversity and yet consists of a number of partially overlapping zones con-
reach their target(s) in the brain. The question then tained in a highly anisotropic lipid bilayer.13 The con-
becomes what physical- and medicinal-chemical char- formational mobility of the lipid chains is relatively low
acteristics do they possess that will enhance the design at or near the water (blood)/lipid interface and increases
of new therapeutic agents at the chemistry and biology strongly toward the interface at the center of the bilayer.
intersection. This article will focus on successful small The lipid-water interface is associated with a layer of
molecule CNS therapeutics and their analogs, empha- perturbed water molecules with significantly different
sizing their physical chemical and pharmacokinetic polarization properties. Because of this, the ability of
properties and characteristics that enable them to pas- these water molecules to form hydrogen bonds with drug
sively cross the BBB. We acknowledge that active molecules is dramatically reduced and forms part of the
transport and efflux are additional important mecha- desolvation process. In addition, the hydrophilic/li-
nisms for transport in the CNS. pophilic interface at the blood/membrane boundary con-
sists of perturbed and bound water, charged polar lipid
head moieties connected to long lipid chains. As a result,
BLOOD-BRAIN BARRIER
a drug approaching the BBB is confronted with a thick
The interface between the blood and an organ is me- layer that is capable of noncovalent interactions with the
diated by endothelial cells that control the transfer of, drug, similarly to that of receptor but with much looser
inter alia, drugs from the blood stream into the cell and steric requirements.
also from the cell to the blood. Basically, there are three The majority of drug BBB penetration is through pas-
methods of transfer for small molecules that can be clas- sive diffusion through the cellular membrane. How this
sified according to whether there is an energy-requiring is accomplished has been the subject of significant re-
step. Two of these processes, passive and facilitated dif- search.
fusion, are concentration driven and unidirectional ac-
cording to the gradient.8 Facilitative diffusion is rela-
ABSORPTION, DISTRIBUTION, METABOLISM,
tively uncommon, but glucose is transferred into the
AND EXCRETION PROPERTIES OF CNS
CNS by a non-energy-dependent glucose transporter.
DRUGS
The third of these processes, active transfer, requires an
energy source (ATP) and can transfer molecules via a For any drug or candidate to achieve optimum thera-
carrier against a gradient. peutic efficacy, it must possess a high degree of potency
In a global sense, the general requirements for passive and selectivity for interaction with a biological target as
and facilitated transport are common. However, depend- well as ability to attain target tissue concentrations that
ing on the organ, the epithelial cells can have differing are above a certain threshold value. Absorption, distri-
specifics for allowing drugs to transfer from the blood to bution, metabolism, and excretion (ADME) processes
the cells in the organ. The CNS, being exquisitely sen- play a pivotal role in defining the disposition of a drug
sitive to many compounds in the blood and also to drugs, candidate, and thus its therapeutic efficacy. Optimizing
is designed to be very selective in what is allowed in. As the chemical structure of lead candidates with respect to
such, it is certainly an outlier as is demonstrated by the the ADME processes has become an integral part of the
lack of correlation between intestinal Caco-2 cell and drug discovery paradigm.14
BBB active efflux.9 An important ADME characteristic is simply the sol-
The reason the BBB is an outlier is that BBB epithelial ubility of the drug as only the amount of drug in solution
cells form tight junctions that effectively preclude para- is available for intestinal absorption and blood distribu-
cellular diffusion. In addition, the cells possess few pi- tion.15 Because diffusion across the BBB is a kinetic
nocytotic vesicles and lack fenestration. Therefore, BBB process and the blood and brain concentrations are in
transfer is through transcellular diffusion through the equilibrium, a concentration of drug in the blood high
membranes. A drug undergoing transcellular diffusion enough to produce a concentration of drug at its receptor
can be metabolized by a formidable battery of metabolic in the brain is critical.16 A drug molecule’s solubility is
enzymes. For example, decarboxylation of 3-(3,4-dihy- a function of many of the physical chemical properties
droxyphenyl)-alanine to dopamine occurs during transit. that we will discuss further on in relation to BBB pene-
An orally active CNS drug needs not only sufficient tration. In addition, not only their bulk properties but also
metabolic stability to maintain integrity in the intestine the placement of the polar and nonpolar areas of the
and liver but also across the BBB. Alternatively, a drug molecule and how those regions influence the inter- and
can be pumped back into the blood by an active transfer intramolecular forces in the crystalline state are relevant
process, mainly through p-glycoprotein, using an ATP to solubility and BBB penetration because water has to
efflux mechanism.10 –12 break these interactions to bring the molecule into solu-

NeuroRx威, Vol. 2, No. 4, 2005

 SUCCESSFUL CNS DRUGS 543

tion. General guidelines for solubility are classified as: genomics) led to the National Institutes of Health (NIH)
moderate at 10 – 60 ␮g/ml and high over 60 ␮g/ml. In the Road Map project.26 The NIH effort is designed to gen-
blood, drugs reversibly bind to proteins like albumin and erate a set of drug-like molecules that are targeted to
␣-acidic glycoprotein.17 The amount of protein binding probe the intersection of chemistry and biology space,
of CNS drugs is an important consideration. The protein albeit dilutely.27 The growth of computing power al-
binding of CNS drugs tends to be rather high. A question lowed the development of in silico approaches that are
that needs to be answered is: as these proteins, carrying based on laboratory-derived data sets. Although ADME
the drugs molecules, approach the polar surface of the covers a host of activities, the remainder of this article
BBB, does the interaction cause the protein to release the will concentrate on the first and major barrier to drug
bound drug? activity, how drugs cross the BBB, and what their phys-
Physiochemical and biological properties are funda- ical chemical and pharmacokinetic properties and char-
mental properties of ADME. Early assessment of the acteristics are.28
physiochemical properties of potential CNS drugs for
their ability to cross the BBB is extremely important.
QSAR STUDIES
Retrospectively, analysis of experimental data has provided
guidelines for physical properties and has been used to The QSAR studies on CNS active drugs and their
derive computational algorithms to predict CNS efficacy. analogs have been summarized through 1988.18 Essen-
The initial analyses of ADME properties, e.g. anesthetic tially, the large number of studies primarily used the
agents in the late nineteenth century, focused on the parti- Hansch approach where the observed biological activity
tion coefficient (LogP) between water and oil, basically the is correlated by variation in the physical properties and/or
lipophilicity of the compound. This has served as one of the structural properties among the molecules in the analog set.
fundamental principles for drug discovery and design.18 There are a large number of descriptors that have been
However, lipophilicity is a two-edged sword. Many other evaluated in the derivation of QSAR equations. The most
drug parameters are affected by lipophilicity. High lipophi- important of these are the physical-chemical properties de-
licity frequently leads to compounds with high rapid met- scribing electronic characteristics, steric effects, solvent
abolic turnover19 and low solubility and poor absorption. partitioning (LogP), and in a much smaller number of cases,
As lipophilicity (LogP) increases, there is an increased molecular weight. The most important of these parameters
probability of binding to hydrophobic protein targets other is lipophilicity that, in many cases, like anesthetics, barbi-
than the desired one, and therefore, there is more potential turates, benzodiazepines, etc., describes the biological ac-
for toxicity. For example, the active sites for undesirable tivity. LogP in all the cases follows a parabolic curve that
cytochrome P450 inhibition and human ether-ago-go- reflects the ability of the molecule to exist in relative pro-
related gene (hERG) K-channel blockade are hydrophobic portions both in aqueous (blood) and lipid membrane en-
and bind lipophilic substrates.20 vironments. Among the electronic effects, the more impor-
Whereas LogP is useful for prediction and optimiza- tant are hydrogen bonding ability, dipole-dipole
tion, it is clear that other physical-chemical properties are interactions, and charge transfer phenomena. By far, the
also of potential importance. The next major advance most important is hydrogen bonding ability.
was due to Hansch and colleagues,21 who developed These early studies underlined the importance of LogP
regression analyses for these properties and initiated the in understanding the activity of CNS drugs and their
field of quantitative structure activity relationships ability to penetrate the BBB. However, these QSAR
(QSAR). Hansch’s initial work was based on a large set equations were designed to both understand the biolog-
of sedative-hypnotic barbituates where he demonstrated ical activity of drugs as well as guide synthesis of novel
biological activity was almost entirely due to their Log P analogs. As such, they were limited in scope and the
and their rate of metabolism was linearly related to LogP. data, and the conclusions derived from were not used to
Furthermore, optimal activity is observed at LogP ⫽ 2.22 make semiquantitative hypotheses about general rules
Other ADME parameters, including cytochrome P450 and molecular requirements for BBB penetration beyond
(CYP450) conversion and drug duration of action, could general observations of lipophilicity.29 In an interesting
also be analyzed using QSAR techniques,18 although it study that proved prescient, Levin determined that rat
was subsequently acknowledged that, for accuracy, data brain permeability was determined by LogP and had a
sets for individual CYPs needed to be analyzed.23 molecular weight cutoff of 400 or less.30 In another,
As the pharmaceutical industry evolved and HTS and Young et al.31 investigated the effect of dipole moment
combinatorial chemistry with their huge chemical librar- and LogP in a series of centrally acting H2-histamine
ies became the norm, the demands for qualification of antagonists related to cimetidine. They proposed that,
molecules with drug-like profiles, many of them virtual, because the molecules were very polar, LogP actually
quickly followed.24,25 The design and use of molecular describes a desolvation effect where the molecules in
arrays as probes for biological function (chemical their hydration spheres need to undergo desolvation at

NeuroRx威, Vol. 2, No. 4, 2005

544 PAJOUHESH AND LENZ

the receptor to realize the hydrogen bonding and dipole 1. Accuracy is the key attribute. In particular, it
interactions with the receptor, in general terms, the BBB should have a minimal number of false negatives.
exterior membrane. 2. Interpretability by medicinal chemists and other
A variation of LogP was introduced by Young et al.,32 related scientists to answer the questions of what
who determined that partitioning of a series of H2-hista- compound is to be made next all the way to what
mine antagonists was best described by a combination of compounds (real or virtual) should be in an opti-
a combination of partition coefficients. The data were mization or screening library.
explained by the difference between the partition coef-
3. Speed is obvious, but there is a trade-off in the
ficients as measured in octanol/water and cyclohexane
execution and accuracy, although this gap is nar-
water (⌬LogP). Octanol, because of its hydroxyl group,
rowing as computing power increases.
is capable of hydrogen bonding, whereas cyclohexane,
being a pure hydrocarbon (lipid) is not. They concluded 4. Robustness is a vital consideration as computa-
that the difference between the two measures ⌬LogP is a tional programs become widely disseminated into a
measure of the hydrogen bonding ability of the molecule. much wider circle of scientists beyond the core
However, the generality of this relationship is unclear as experts.
a similar study on a series of H1 antagonists only showed The data set for much of the original research on BBB
dependence on LogP.33 Subsequently, Goodwin et al.34 penetration is based on a set of more than 100 com-
indicated that LogP is predominantly a measure of drug pounds whose LogBBB was determined by radiolabeled
volume, or surface area, plus hydrogen bond acceptor drug determinations in blood and brain of anesthetized
potential. Thus, both hydrogen bonding potential and rats.40 This is a slow, labor intensive, and costly process
drug volume contribute to permeability. These general but has the highest predictability. The structures of these
and cited studies, formed the basis for elucidation of the compounds are illustrated in a review. Although these
mechanism of BBB penetration by drugs that followed. LogP values are for the BBB, it should be mentioned that
Because most CNS drugs have either basic (mostly) or calculation of general LogP values probably has the
acidic properties, or are amphiphilic, molecular charge highest accuracy among computational chemistry pro-
should be a component of BBB penetration attributes. grams, whereas calculation of LogBBB is of more lim-
Because these drugs are weak bases or acids (compared ited reliability.41
with strong fully ionized ones like sulfuric acid and sodium Much larger data sets, on the order of many thousands
hydroxide) as measured by their pKa (log of the acid ion- of compounds, have used a classification system
ization constant), they are in equilibrium with their un- whereby compounds are classified as CNS active, inac-
charged neutral species. In measuring LogP, the species in tive, or questionable. Generally, these databases are the
the water layer may be an equilibrium mixture of charged World Drug Index (WDI), Comprehensive Medicinal
and neutral species. However, the species that is extracted Chemistry database (CMC), the MDL Drug Data Report
and exists in the organic layer is neutral.35 There are rare (MDDR), and the Available Chemicals Directory
exceptions to this, like glycine.36 So, in effect, what is being (ACD). As Clark39 has pointed out, the CNS active drugs
measured is a distribution ratio. The unraveling of LogP are readily identifiable (although active transport may be
with charged species as a function of pH is complicated. In involved), whereas the CNS inactive are more difficult to
effect, the amount of uncharged neutral species in water is identify. This is because the drugs may actually penetrate
critical for membrane penetration. Fischer et al.,37 in a study the CNS but are either removed by efflux mechanisms or
of comparative properties for BBB penetration, estimated may be metabolized to inactive compounds.
pKa limits for penetration between 4 and 10. Similar con-
clusions were made by Palm et al.38 but using intestinal
COMPUTATIONAL QSAR
absorption in Caco-2 cells.
Through the Hansch QSAR equations, lipophilicity, The initial computational approach to BBB penetra-
hydrogen bonding, and molecular weight have been tion by computing LogBBB was published by van de
identified as being generally critical to BBB penetration. Waterbeemd and Kansy.42 Their data set was based on a
However, with different parameters these qualities are subset of the compounds studied by Young et al.32 Al-
also applicable to membrane transit in other types of though molecules are flexible, the study used the lowest
cells in other organs. energy conformation for the calculation and required
construction of a three dimensional conformationally
minimized structure for each molecule. As it turned out,
COMPUTATIONAL APPROACHES TO BBB
based on later work, this was justified. The polar surface
PENETRATION
area (PSA) and the molecular volume components were
For a computational (in silico) method to be useful, it the most important descriptors, with PSA strongly pre-
needs to have four attributes39: dominating.42 PSA is defined as the surface area (Å2)

NeuroRx威, Vol. 2, No. 4, 2005

 SUCCESSFUL CNS DRUGS 545

occupied by nitrogen and oxygen atoms and the polar the same small data set using a single low-energy con-
hydrogens attached to them and is strongly reflective of formation. Interestingly, using only the calculated PSA
hydrogen bonding capacity and polarity. Both of these did not yield the best results. Adding the calculated LogP
forces are involved as the molecule approaches the polar (ClogP) to the QSAR equation gave more accurate re-
surface of the membrane and desolvates as it moves into sults, essentially adding a lipophilicity term to the hy-
the lipid portion. Although the calculated results were drogen bonding and polarity terms inherent in PSA. The
indicative, the training set as later demonstrated was too advantage to Clark’s QSAR equation is that by using a
small to extrapolate to the general case. single conformer is fast and easy to calculate LogBBB
Abraham constructed a larger set of molecules in an and thus predict whether a molecule may pass the BBB.
effort to reliably calculate LogBBB.43 The QSAR equa- This makes the equation suitable for evaluating large
tion derived contains terms for excess molecular refrac- screening libraries.
tion, solute polarizability, hydrogen bond acidity, and It is reasonable to speculate on the reasons that use of
basicity and molecular volume. Although the agreement single low-energy conformation of a drug yields a useful
between calculation and experiment is good and this estimate of LogBBB and BBB penetration. First, many,
procedure has been used extensively by Abraham and his but not all, of the drugs contained in the small set do not
group, it fails Rule 2 above as medicinal chemists will have a great deal of structural flexibility. This limits the
find it difficult in using these types of descriptors in set of low energy conformations to a relatively compact
designing compounds. volume. Part of this is the consequence of the optimiza-
Because the initial PSA research used the low energy tion process that a medicinal chemist undertakes during
conformation of each molecule and most molecules have discovery where rigidity is incorporated into the mole-
many accessible conformations at body temperature,
cule to increase biological activity among other factors.
Palm et al.44 developed a dynamic PSA approach
Second, the BBB acts as a molecular filter and not a
whereby the set of available conformations were used
barrier. Some compounds need to be let in and others
and the contribution of each to the overall PSA was
kept out. There is a molecular recognition event when a
calculated using a Boltzman distribution thereby taking
drug approaches the external polar face of the BBB.
into account conformational flexibility. Based on their
Given the diverse nature of the CNS drugs, this interac-
results in intestinal Caco-2 cells, drugs with a PSA of 60
tion is more of a physical chemical nature than steric, i.e.
Å2 or less are completely absorbed, whereas those with
it does not have the stricter steric requirements that a
at least 140 Å2 are not. This approach is computationally
intensive and not suitable for analyzing large libraries. receptor-ligand pair would have.
As it turned out, other investigators demonstrated that Because of the necessity of screening very large real or
a single low-energy conformation is sufficient for deter- virtual libraries, on the order of a million compounds or
mining PSA. Kelder estimated LogBBB using a set of 45 more, approaches were investigated that would signifi-
compounds, mostly the same set as others used, and cantly speed the evaluation of each compound while
compared PSAs for both single lowest energy conforma- maintaining the accuracy. Ertl has developed a topolog-
tions as well as a Boltzman average of the 100 lowest ical PSA (TPSA) approach that fits these criteria.48
energy conformations.45 There was essentially no differ- TPSA is based on dissecting the contributions of polar
ence between the two approaches, indicating the compu- groups in drugs contained in the WDI. Comparison with
tationally intensive conformational calculations are not Clark’s results demonstrated almost no difference be-
required. They further used these compounds as a train- tween the two approaches. The major advantage of
ing set and compared a set of 776 orally active CNS TPSA is that it uses a two-dimensional structure and not
drugs that had entered phase II clinical trials against a set the computationally more intensive three-dimensional
of 1590 non-CNS drugs that had also entered phase II. representation. This enables two to three orders of mag-
Using only a single low-energy conformation of each nitude increases in throughput.
drug for the PSA calculation gave good agreement and Because PSA is dependent upon hydrogen bonding
separation based on PSA between the CNS and non-CNS and donating atoms, Österberg49 investigated simplify-
drugs. Kelder found that non-CNS drugs transported pas- ing the PSA calculation by incorporating hydrogen bond-
sively and transcellularly needed a PSA of 120 Å2 or ing descriptors in place of PSA while including ClogP.
less, whereas the drugs can be targeted to the CNS with There was high correlation between the two methods,
a PSA less than 60 –70 Å2. Similar conclusions were indicating the importance of hydrogen bonding and li-
made by van de Waterbeemd based on a study of mar- pophilicity for LogBBB. As with TPSA, this simplifica-
keted CNS and non-CNS drugs.46 Their cutoff for PSA tion allows BBB penetration evaluation of large actual
cutoff for CNS penetration is 90 Å2 or below and a and virtual libraries. This hydrogen bonding approach
molecular weight cutoff of 450. was later extended and developed into a pair of rather
Clark47 also used PSA as a measure of LogBBB using simple rules for predicting BBB penetration:

NeuroRx威, Vol. 2, No. 4, 2005

546 PAJOUHESH AND LENZ

FIG. 1. Molecular dynamics simulations of acyclovir and diazepam passive transport through a DMPC membrane. (Courtesy of Prof.
A. J. Hopfinger, University of Illinois-Chicago, Chicago, IL. Copyright 姝 2005, The Chem21 Group, Inc., Chicago, IL. All rights reserved.).
Top: Side view (A) and top view (B) of the structure for the model DMPC membrane-acycloviar complex after 200 psec of MDS. Acyclovir
has low (slow) cellular permeability Bottom: Side view (A) and top view (B) of the structure for the model DMPC membrane-diazepam
complex after 200 psec of MDS. Diazepam has high (rapid) cellular permeability.

1. Because hydrogen bonding is primarily associated penetrator) and diazepam (a fast penetrator) going
with oxygen and nitrogen moieties in a molecule, through an artificial membrane DMPC that corresponds
then, if the sum of the nitrogen (N) and oxygen (O) closely with biological membranes. Acyclovir has just
atoms in the molecule is five or less, then the mol- passed the polar end while diazepam is almost through
ecule has a high probability of entering the CNS. half of the bilayer. The illustration demonstrates the path
the drugs need to follow from both the side and top
BBB penetration ⫽ (N ⫹ O) ⫽ ⱕ5
views. It is important to realize that the membrane is
2. If ClogP –(N ⫹ O) ⬎ 0, then LogBBB is likely to fluid and especially the lipid side chains have large de-
be positive and the compound has a high probabil- grees of freedom to move and interact and thus retard
ity of entering the CNS. penetration by the drugs. PSA and ClogP were, as before,
The sum of the (N ⫹ O) atoms actually measures the the most important descriptors but, added to these, were
hydrogen bond acceptors associated with nitrogen and descriptors for the interaction between the drug and the
oxygen moieties. It should be pointed out that there are phospholipid polar head and drug molecular flexibility.
other heteroatoms in drugs that can function as hydrogen The polar interaction descriptor recognizes the molecular
bond acceptors (HBA) and total HBA, including (N ⫹ recognition event at the start of the penetration, whereas
O) would probably give a better measure. the flexibility descriptor corresponds to the ability of a
A molecular recognition event occurs between a drug drug to essentially wiggle through the array of long chain
and the polar end of the phospholipid membrane at the lipids in the membrane. Veber et al.51 have found, how-
initiation of BBB penetration. This type of interaction ever, that increasing flexibility hinders membrane transit.
was explicitly incorporated in a membrane interaction It is likely, however, that the amount of molecular flex-
QSAR (MI-QSAR). In addition to the descriptors used in ibility will follow a parabolic curve similar to LogP
studies above, Hopfinger and colleagues50 used a molec- where some amount facilitates transit through the mem-
ular dynamics trajectory simulation of the drugs through brane, but too much conformational mobility interferes
a synthetic membrane. Figure 1 shows acyclovir, a slow with the fluid lipid side chains.

NeuroRx威, Vol. 2, No. 4, 2005

 SUCCESSFUL CNS DRUGS 547

Hopfinger and colleagues52 used another approach to through the blood brain barrier, when the molecular mass
BBB penetration in developing QSAR equations across is kept in or below a 400- to 600-Da range.57 For mar-
diverse sets of molecules based on a 150-compound data keted CNS drugs, the mean value of MW is 310, com-
set. This QSAR approach uses cluster analyses employ- pared with a mean MW of 377 for all marketed orally
ing four-dimensional molecular similarities to partition active drug.53
compounds into clusters, and membrane-interaction MI-
QSAR analysis to build models for each cluster. In ef- Hydrogen bonding
fect, this means identifying and using common pharma- All the QSAR equations emphasize the importance of
cophores to group large numbers of diverse molecules hydrogen bonding whether through polarity, PSA, hy-
into sets and building QSARs based on each set with the drogen bond donor and acceptor counts, or simply count-
objective of building optimum QSAR equations for each. ing heteroatoms capable of hydrogen bonding. All of
Provided it is computationally tractable to identify these these measurements are correlated, for instance, (O ⫹ N)
pharmacophores in this manner, this technique should be atom count is highly correlated with PSA but measures
of use in screening large libraries as it potentially con- hydrogen bond acceptors. CNS penetration requires a
tains a basis for generating relevant screening models for
sum of these heteroatoms of 5 or less.49 Compounds with
wide chemical diversity.
high hydrogen bond forming potential, such as peptides
with their amide groups, peptides even as small as di- or
QSAR SUMMARY tripeptides, have minimal distribution through the
Examination of the QSAR descriptors that have ap- BBB.11 Increasing hydrogen bonding decreases BBB
peared in the various equations provides insight into the penetration. The marketed CNS drugs on average have
important molecular factors that govern BBB penetra- (O ⫹ N) ⫽ 4.32, PSA ⫽ 16.3%, 2.12 hydrogen bond
tions. Importantly, because these studies appeared before acceptors and 1.5 hydrogen bond donors.53 This study
a summary of the same properties calculated from all the indicates that polar properties of CNS drug are reduced
orally active drugs, numbering 74, approved since 1983 relative to other parameters.
for CNS use appeared, the underlying validity of the
conclusion and their molecular implications can be Polar surface area
made.53 PSA has been used as a predictor for BBB penetration
by many investigators.6,58 In general, drugs aimed at the
Lipophilicity CNS tend to have lower polar surface areas than other
Lipophilicity was the first of the descriptors to be classes.59,60 PSA for CNS drugs is significantly less than
identified as important for CNS penetration. Hansch and for other therapeutics with PSA for CNS penetration
Leo54 reasoned that highly lipophilic molecules will par- estimated at 60 –70 Å2 through 90 Å2.45,46 The upper
titioned into the lipid interior of membranes and will be limit for PSA for a molecule to penetrate the brain is
retained there. However, ClogP correlates nicely with around 90 Å2.
LogBBB with increasing lipophilicity increasing brain
penetration. For several classes of CNS active sub- Molecular volume and flexibility
stances, Hansch and Leo54 found that blood-brain barrier These descriptors both need further research. Molecu-
penetration is optimal when the LogP values are in the lar volume is a function of MW and structure and takes
range of 1.5-2.7, with the mean value of 2.1. An analysis into account all the accessible conformations available to
of small drug-like molecules suggested that for better the molecule under physiological conditions. This actu-
brain permeation46 and for good intestinal permeability55 ally relates to rotatable bonds and the number of rings in
the LogD values need to be greater than 0 and less than the molecule. Rotatable bond count is now a widely used
3. In comparison, the mean value for ClogP for the filter following the finding that greater than ten rotatable
marketed CNS drugs is 2.5, which is in good agreement bonds correlates with decreased rat oral bioavailability.51
with the range found by Hansch et al.22 CNS drugs have significantly fewer rotatable bonds than
Molecular weight other drug classes. Most centrally acting compounds
CNS drugs have significantly reduced molecular have rotatable bond count of five or less.53 Based on the
weights (MW) compared with other therapeutics. single conformation PSA calculations, it appears that the
Levin30 has a cutoff for BBB penetration of 400. Van de conformational range is limited in CNS drugs. Molecular
Waterbeemd56 has suggested that MW should be kept flexibility relates to the ease by which the molecule
below 450 to facilitate brain penetration and to be lower transverses the membrane, and it appears that limited
than that for oral absorption.46 The rules for molecular flexibility is advantageous but significant is disadvanta-
weight have been reviewed, where small molecules may geous.52 More research is needed to determine more
undergo significant passive lipid-mediated transport general and quantitative measurements.

NeuroRx威, Vol. 2, No. 4, 2005

548 PAJOUHESH AND LENZ

Charge CYP1A2, and CYP2C19. Inhibition of CYP enzymes is

Many CNS drugs are basic and exist in an equilibrium the most common cause of drug interactions. Commonly
between their charged and neutral states under physio- a concentration of 10 or 30 ␮M of drug that inhibits the
logical conditions or are amphiphilic if they also possess metabolic activity of a given enzyme by less than 50% is
an acidic group. Possession of a positive charge at pH considered to have an acceptable level of CYP inhibition,
7– 8 tends to favor brain permeation.61 Additionally, although the less the better. To maximize oral absorp-
compounds possessing a tertiary nitrogen (a feature of tion, a successful CNS drug must have no significant
many CNS drugs) show a higher degree of brain perme- CYP2D6 metabolism and be a nonpotent CYP3A4 in-
ation. As molecules partition into membrane lipids as ducer. CYP3A4 and 2D6 are of great interest because of
neutral species, this will be a function of both the con- their high probability to be involved in the metabolism of
centration of the neutral species as well as its lipophilic many coadministered drugs. A compound that is potent
properties. Strong bases and acids are thereby pretty inhibitor of one of these two enzymes has a good chance
much precluded from BBB penetration, which accords of inhibiting the metabolism of coadminstered drugs,
well with the known difficulty that carboxylic acids have causing their plasma levels to be elevated, possibly into
in penetrating the CNS.62 Fischer et al.37 put pKa limits toxic ranges. In fact, mibefradil, the first T-type calcium
on BBB penetration of between 4 and 10. channel blocker, was introduced to the market in 1997,
and then abruptly withdrawn.65 Pharmacokinetic interac-
Pharmacokinetic properties tions with other drugs metabolized by P450, CYP3A4,
A potential drug’s pharmacokinetic properties to a sig- and 2D6, eventually led to withdrawal of mibefradil from
nificant extent can be manipulated by the medicinal the clinic.
chemist. However, historically, these properties have not Permeability. Oral absorption of drugs is also deter-
been considered by most medicinal chemists, who in mined by their permeability. The relative contribution of
many cases used increases in receptor binding for their the transcellular and paracellular pathways, to overall
optimization strategy. In this section. we briefly look at absorption of a compound is highly dependent on its
the pharmacokinetic properties that impact on ADME physicochemical properties of the compounds. Most rel-
properties of CNS drugs. These parameters include: evant physicochemical properties that affect membrane
permeability and are related to each other have already
• Metabolic stability
been mentioned.66 However, lipophilicity (log D) and
• Metabolic liability MW are the two main components that affect the per-
• Permeability meability the most. Other factors that influence the per-
• Protein binding meability are bioisosteric replacements.67 For example,
replacement of a carboxylic acid with a tetrazole can
• hERG inhibition
improve permeability.
Metabolic stability. First-pass metabolism influences Protein binding. Protein binding is a key factor for
oral bioavailability and toxicology of drugs. The liver BBB penetration. CNS drugs that are normally weakly
and intestines are the major sites of such metabolism. basic molecules bind to both human serum albumin
Many discovery compounds exhibit low bioavailability (HSA) and ␣1-acid glycoprotein (AGP). Depending on
because of high rates of metabolism. Rapid metabolism the drug and the target, high affinity for plasma proteins
is one of the primary reasons for molecules not reaching may consequently be an asset or a drawback for efficacy.
adequate systemic levels. These are phase I cytochrome HSA is the most abundant protein in blood plasma and
P450 oxidations or phase II conjugations. Metabolism serves as a transport protein and considered to have low
reduces the circulating drug concentration and increases affinity and high capacity for binding, whereas AGP has
elimination. Typically, if the percentage remaining of a low capacity and high affinity for binding.68 Drug-pro-
compound at 60 min is less than 20%, the compound tein interaction is a reversible process and a successful
considered to be metabolically unstable. Low metabolic CNS drug should not be an efficient P-glycoprotein sub-
stability results in a high clearance and short half-life of strate (in vivo) and not a high-affinity serum albumin
a compound. The greater than 80% remaining of drug at ligand (Kd ⬍10 ␮M).69
60 min is considered to be ideal.63 hERG inhibition. FDA concern over the risk of life-
Metabolic liability. CYPs constitute the most impor- threatening arrhythmias associated with QT interval pro-
tant family of biotransformation enzymes involved in longing drugs has led to recommendation of more thor-
drug metabolism, playing an important role in the dis- ough preclinical and clinical cardiovascular safety
position of drugs and their pharmacological and toxico- assessments. In particular, most of the QT-prolonging
logicals effects.64 It comprises a superfamily of isoforms drugs have been shown to inhibit the K⫹ channels en-
that are vital in the metabolism of drugs. The most com- coded by the hERG.70 Drugs that interact with Ca2⫹ and
mon used isoforms are CYP3A4, CYP2C9, CYP2D6, Na⫹ channels have a great tendency to block hERG. For

NeuroRx威, Vol. 2, No. 4, 2005

 SUCCESSFUL CNS DRUGS 549

better risk assessment, a more than 30-fold margin be- cal tools for in silico screening of large compound li-
tween hERG IC50 and effective unbound plasma concen- braries for identifying potential leads and early ADME
trations are recommended for drugs in clinical develop- profiling but not for high-resolution physical chemical
ment.71 It needs to be pointed out that challenge is to understanding of BBB penetration.
minimize hERG activity to eliminate a major risk factor, Although the previous comparative analyses are useful
but not at the expense of discarding therapeutically use- for large libraries, Lipinski et al.76 at Pfizer were looking
ful drugs that are urgently required to treat a myriad of for a set of general rules that would govern drug-like
diseases for which there is presently a paucity of quality properties aimed primarily at solubility and membrane
treatments. penetration that could be used by medicinal chemists.
Lipinski et al. noted that with the advent of HTS and
COMPARATIVE CLASSIFICATION METHODS combinatorial chemistry starting in the late 1980s, the
trend for lead molecules identified by HTS was toward
An alternative way of evaluating BBB permeation is increasing molecular weight and increasing lipophilicity.
to conduct retrospective analyses based on large data- Both of these contributed toward decreased solubility
bases containing qualitative data on drugs, such as the and membrane penetration. A database of about 2500
WDI, MDDR, and CMC. The approach is to compare orally active drugs that had entered at least phase II,
attributes across large numbers of drugs to determine based on United States Adopted Name (USAN) or In-
what their common attributes are and what the range of ternational Non-proprietary Name (INN) names, was
these are. used to analyze the physical chemical properties of these
Adenot extracted a large CNS library from the WDI drugs. From this analysis, the “Rule of Five” was devel-
containing approximately 1700 compounds segregated oped. The “Rule of Five” is so named because all the
according to positive and negative BBB penetration and essential physical properties are parameters of five. Ac-
whether the molecule was a Pgp substrate.72 Discrimi- cording to this rule, a good absorption and permeability
nation analysis yielded an initially surprising result. Most is likely if:
of the non-BBB-penetrating compounds have a number
of heteroatoms larger than 8, whereas most of the BBB • Molecular weight is ⱕ500
permeable compounds have a number of heteroatoms • Oil/water distribution coefficient (LogP) is ⱕ5
lower than 9. The reason that the number of heteroatoms • Hydrogen bond donors ⱕ5 (expressed as the sum of
classified most of the compounds is due to its correlation OHs and NHs)
to the polar surface area. In this study, heteroatoms in-
• Hydrogen bond acceptor ⱕ10 (expressed as the
cluded oxygen and nitrogen as well as phosphorus, sul-
sum of Ns and Os)
fur, and halogens, explaining the difference with Öster-
berg’s studies.49 This approach, however, compares two A fifth rule was added later:
sets of drugs looking for differences. As a result, the
• Number of rotatable bonds ⱕ10
similarities are factored out.
Neural networks are being used as part of an effort to This was originally based on rat studies34 but is sup-
qualify compounds based on their “drug-likeness” for ported by the observation that drugs discontinued during
inclusion in large screening libraries. Sadowski and Ku- the clinical development process have more rotatable
binyi73 encoded and analyzed over 200,000 molecules bonds than those that are successfully marketed.77 This
from WDI and ACD using topological descriptors. Neu- effect may very well be a reflection of a conformation-
ral network analysis qualified drug-likeness on non-drug- ally mobile compound trying to permeate through mobile
likeness success percentage of approximately 80%. A lipid chains in the bilayer.
similar study by Ajay74 used a Bayesian neural network These rules were explicitly stated as not applying to
to discriminate between drug-likeness and non-drug- actively transported substrates. If any two of the rules
likeness. Subsequently, Ajay and Murcko75 published a were exceeded, the compounds were likely to possess
study on designing libraries with CNS activity. They poor permeability and solubility properties.78 Because,
reported the development of a neural network trained on however, these properties are all correlated in individual
different sets of about 9000 compounds from CMC and molecules, exceptional strength in one could counteract a
MDDR drug databases separated into CNS-active and substandard score in another.
CNS-inactive drugs. They used an automated classifica- Because Lipinski’s data set contained drugs that en-
tion scheme based on therapeutic use. After training, tered clinical trials, the probability existed that many of
classification of test sets of about 13,000 CNS-active and these failed on the path to approval. Wenlock77 evaluated
53,000 CNS-inactive compounds from both databases a set of marketed drugs and those that failed in clinical
was performed with approximately 80% accuracy. These trials. The results are shown in Table 1. The agreement
neural network-based classifications appear to be practi- between Lipinski’s Rules (clinical candidates set) and

NeuroRx威, Vol. 2, No. 4, 2005

550 PAJOUHESH AND LENZ

TABLE 1. Comparison of Oral Lipinski’s Dataset and TABLE 3. Comparison of All Marketed Oral Drugs
Wenlock’s Marketed Drugs Dataset* between 1983 and 2002 and CNS and Gastrointestinal-
Metabolic Drugs According to Leeson
Physical Chemical Property Lipinski Wenlock
Physical
Molecular weight 500 473 Chemical All Drugs* CNS GI/Metabolic
Calculated LogP 5 (ClogP) 5.5 (ACD LogP) Properties (n ⫽ 329) (n ⫽ 74) (n ⫽ 38)
Calculated LogD7.4 4.3
Hydrogen bond donors 5 4 Molecular weight 377 310 378
Hydrogen bond acceptors 10 7 ClogP 2.50 2.50 1.90
%PSA 21.0 16.3 26.7
*90% Cutoff. (OH ⫹ NH) 1.77 1.50 2.71
(O ⫹ N) 6.33 4.32 6.84
Hydrogen bond 3.74 2.12 4.34
Wenlock’s marketed drug data set is excellent. Vieth79 acceptors
also studied a set of 1193 marketed drugs with results Rotatable bonds 6.42 4.70 7.63
essentially confirming those of Wenlock and Lipinski.80 Aromatic rings 2.88 2.85 2.32
Lipinski80 also laid down a set of rules derived from
the set of 1500 drugs that were filtered from USAN or *All drugs contains the 74 CNS drugs as part of the data set.
INN names for good CNS penetration. CNS penetration
is likely if: results (Table 2) did not differentiate between the two
classes in molecular weight, volume, hydrogen bond ac-
• Molecular weight ⱕ400 ceptors, and number of aromatic rings. Significant dif-
• Log p ⱕ 5 ferences were observed with hydrogen bond donors,
• Hydrogen bond donor ⱕ3 ClogP, PSA, and molecular flexibility (rotatable bonds).
In general, although there are some differences, perhaps
• Hydrogen bond acceptor ⱕ7
due to the smaller data set, the results are in agreement
This study indicated that, for CNS penetration, the with both laboratory and computational results.
physical properties in general have a smaller range than An interesting study was published by Leeson and
general therapeutics. These “Rules” certainly confirmed Davis53 that used the year of market introduction of
the observations made independently by earlier workers. drugs to determine not only their physical chemical prop-
Lipinski’s contribution is that the study was based on erties but also how they may have changed as a function
clinically studied drugs and looked at their overall re- of time. The time course was grouped into two parts,
quirements not against a set of control drugs where com- launched before 1983 (864 drugs) and launched between
mon factors were filtered out. Lipinski’s Rule of Five has 1983 and 2002 (329 drugs). The pre-1983 was taken
thus provided medicinal chemists with a simple mne- from Vieth’s79 compilation and from 1983 from the last
monic for identifying compounds with medicinally rele- chapter in Annual Reports in Medicinal Chemistry. In
vant physical chemical properties. As such, medicinal contrast to all the other studies, Leeson’s was subdivided
chemists have enthusiastically adopted the Rule of into six therapeutic areas, i.e. CNS, cardiovascular, anti-
Five.81 infective, etc. This study demonstrated that cardiovascu-
Two studies have contrasted physical chemical prop- lar and CNS drugs were both outliers compared to the
erties of CNS-marketed drugs to either non-CNS drugs other classes (Table 3). For the 74 CNS drugs, the results
in general or to other classes of non-CNS drugs. Mahar show, compared with all drugs, that these drugs have
Doan82 conducted both biochemical and computational statistically different (␳ ⫽ 0.001 ⫺ 0.01) means and
experiments on 48 CNS and 45 non-CNS drugs. The medians in some physical chemical aspects and do not

TABLE 2. Mean (Range) of Physical Chemical Properties of CNS and Non-CNS Drugs from Mahar Doan

Physical Chemical Properties CNS Non-CNS

Molecular weight 319 (151–655) 330 (163–671)

ClogP 3.43* (0.16–6.59) 2.78* (⫺2.81–6.09)
ClogD 2.08 (⫺1.34–6.57) 1.07 (⫺2.81–5.53)
PSA 40.5 (4.63–108) 56.1 (3.25–151)
Hydrogen bond donors 0.85* (0–3) 1.56* (0–6)
Hydrogen bond acceptors 3.56 (1–10) 4.51 (1–11)
Flexibility (rotatable bonds) 1.27* (0–5) 2.18* (0–4)
Aromatic rings 1.92 (0–4) 1.93 (0–4)
*Statistically different.

NeuroRx威, Vol. 2, No. 4, 2005

 SUCCESSFUL CNS DRUGS 551

TABLE 4. Attributes of a Successful CNS Drug less polar, have less flexibility, and have lower molecular
weights than drugs used for other therapeutic indications.
● Potent activity: low to subnanomolar In addition, the molecular volume of CNS drugs appears
● Highly selective
● Molecular weight ⬍ 450 to be smaller than other drugs.
● Minimal hydrophobicity (clogp ⬍ 5) In terms of design of new drugs or analogs, besides the
● Number of H-bond donor ⬍ 3 physical chemical attributes listed above, the fact that
● Number of H-bond acceptor ⬍ 7 these attributes are correlated is important. Thus, chang-
● Number of rotatable bonds ⬍ 8 ing a molecule to adjust one attribute will change others
● H-bonds ⬍ 8
● Pka, neutral or basic with pKa 7.5–10.5 (avoid acids) as well. This was clearly shown by a simple QSAR
● Polar surface area ⬍ 60–70 Å2 equation where LogP is clearly defined by molecular
● ⬎30-Fold margin between hERG IC50 and effective weight and hydrogen bond donors and acceptors. In de-
unbound plasma concentration. signing new drugs, there is a balancing act required to
● Metabolic stability with ⬎80% remaining after 1 h. address the physical chemical requirements and making
● P450 enzyme CYP inhibition ⬍ 50% at 30 ␮M.
● No significant CYP2D6 metabolism. the best compromises in properties from those available
● Not a potent CYP3A4 inducer. and the actuator requirements at the receptor. At the end,
● Not an efficient P-glycoprotein substrate (in vivo). attributes of a successful CNS drug are shown in Table 4
● Not a high-affinity serum albumin ligand (Kd ⬍ 10 and can be used as a guide to design CNS therapeutic
␮M). agents with better drug-like properties.
● Aqueous solubility ⬎ 60 ␮g/ml.
● Effective permeability ⬎1 ⫻ 10⫺6cm/sec As this article focused on passive transport of drugs
into the CNS, other factors like active transport, efflux,
and metabolism need to be addressed and unraveled be-
fore a potential CNS drug can enter the clinical devel-
differ in others. They have significantly reduced molec- opment pathway. The entire area of active transport pro-
ular weight, polar properties %PSA [(O⫹N), total hy- teins is undergoing intensive research, and their steric
drogen bond acceptors], (OH ⫹ NH) hydrogen bond and energetic requirements will be the next major area to
donors), and rotatable bonds (molecular flexibility). aid in both design of new CNS drugs and the develop-
Thus, indicating that CNS drugs are smaller with a more ment of new and safer agents.
compact and less flexible structure, and that the surface
has fewer polar groups able to function as hydrogen bond
donors and acceptors and that compared to the total REFERENCES
surface area, the PSA is reduced.
1. Pardridge WM. The blood-brain barrier and neurotherapeutics.
Interestingly, there is little difference between the NeuroRx 2:1–2, 2005.
physical chemical properties of all drugs and the number 2. Mouritsen OG, Jorgensen K. A new look at lipid membrane struc-
of rings and lipophilicity as measured by ClogP. The ture in relation to drug research. Pharm Res 15:1507–1519, 1998.
3. Lin JH, Lu A-Y. Role of pharmacokinetics and metabolism in drug
average ClogP for CNS drugs is 2.50, which is close to discovery and development. Pharmacol Rev 49:403– 449, 1997.
that of cardiovascular and gastrointestinal drugs but sig- 4. Bannon WW, Deceker MW, Holladay MW, Curzon P, Donelly-
nificantly different from those of anti-infective, respira- Roberts D, Puttfarcken PS, Bitner RS, Diaz A, Dickenson AH,
Porsolt RD, Williams M, Arneric SP. Broad spectrum, non-opioid
tory, and inflammation, and cancer therapeutics. analgesic activity by selective modulation of neuronal nicotinic
acetylcholine receptors. Science 279:77– 81, 1998.
5. Daly JW, Myers CW, Whittaker N. Further classification of skin
SUMMARY alkaloids from neotropical poison frogs (Dendrobatidae), with a
general survey of toxic/noxious substances in the amphia. Toxicon
The requirements for passive CNS penetration by 25:1023–1095, 1987.
drugs is reasonably well understood, but additional work 6. Lenz GR. Technical problems in getting results. In: From data to
drugs: strategies for benefiting from the new drug discovery tech-
still needs to be done on the effect and optimal range of nologies (Haberman AB, Lenz GR, Vaccaro DE, eds.), pp 95–114.
molecular flexibility on penetration. This will probably London: Scrip Reports, 1999.
occur through additional molecular dynamics simulation 7. Williams M, Coyle JT, Shaikh S, Decker MW. Same brain, new
decade: challenges in CNS drug discovery in the postgenomic,
of membrane transit. proteomic era. Annu Rep Med Chem 36:1–10, 2001.
To function as a drug, a molecule has to meet a set of 8. Schneeberger EE, Lynch RD. Structure, function, and regulation of
physical chemical requirements in addition to the steric cellular tight junctions. Am J Physiol (Lond) 262:L647–L661,
1992.
and energetic requirements at its receptor. Many of these
9. Faasen F, Vogel G, Spanings H, Vromans H. Caco-2 permeability,
requirements are the same for both CNS and non-CNS P-glycoprotein transport ratios and brain penetration of heterocy-
drugs. For instance, whereas lipophilicity is important clic drugs. Int J Pharm 263:113–122, 2003.
for CNS penetration, the ClogP average values for CNS 10. Graff CL, Pollack GM. Drug transport at the blood-brain barrier
and the choroid plexus. Curr Drug Metab 5:95–108, 2004.
and cardiovascular drugs are the same. Based on the data 11. Pardridge WM. CNS drug design based on principles of blood-
in the tables, CNS drugs tend to be more lipophilic, be brain barrier transport. J Neurochem 70:1781–1792, 1998.

NeuroRx威, Vol. 2, No. 4, 2005

552 PAJOUHESH AND LENZ

12. Schlosshauer B, Steuer H. Comparative anatomy, physiology and medicinal chemistry. Vol. 4 (Sammes PG, Taylor JB, eds.), pp
in vitro models of the blood-brain and blood-retina barrier. Curr 241–294. Oxford, UK: Pergamon, 1990.
Med Chem - Central Nervous System Agents 2:175–186, 2002. 36. Taylor PJ. Hydrophobic properties of drugs. In: Comprehensive
13. Sippl W. Computational approaches for the prediction of blood- medicinal chemistry. Vol. 4 (Sammes PG, Taylor JB, eds.), p 263.
brain barrier permeation. Curr Med Chem –Central Nervous Sys- Oxford, UK: Pergamon, 1990.
tem Agents, 2:211–227, 2002. 37. Fischer H, Gottschlich R, Seelig A. Blood-brain barrier perme-
14. Dennis M. Absorption processes, In: Comprehensive medicinal ation: molecular parameters governing passive diffusion. J Mem-
chemistry. Vol. 5 (Sammes PG, Taylor JB, eds.), pp 1– 44. Oxford, brane Biol 165:201–211, 1998.
UK: Pergamon, 1990. 38. Palm K, Luthman K, Ros J, Gråsjo J, Artursson P. Effect of
15. Kerns EH, Di L. Pharmaceutical profiling in drug discovery. Drug molecular charge on intestinal epithelial drug transport: pH-depen-
Discov Today 8:316 –323, 2003. dent transport of cationic drugs. J Pharmacol Exp Ther 291:435–
16. Faller B, Wohnsland F. Physicochemical parameters as tool in drug 443, 1999.
discovery and lead optimization. In: Testa B, Van de Waterbeemd 39. Clark DE. Prediction of intestinal absorption and blood-brain bar-
H, Folker G, Guy R. Pharmacokinetics optimization in drug re- rier penetration by computational methods. Comb Chem High
search. Zurich, Switzerland: Wiley-VCH; 257–273, 2001. Throughput Screening 4:477– 496, 2001.
17. Ellingboe J. Application of statistical analysis techniques for dis- 40. Ecker GF, Noe CR. In silico prediction models for blood-brain
covery compound property optimization. AAPS Workshop on Op- barrier permeation. Curr Med Chem 11:1617–1628, 2004.
timization of Drug-Like Properties during Lead Optimization. Par- 41. Jorgenson WL. The many roles of computation in drug discovery.
sippany, NJ, 19 –22 September, 2004. Science 303:1813–1818, 2004.
18. Gupta SP. QSAR studies on drugs acting at the central nervous 42. van de Waterbeemd H, Kansy M. Hydrogen bonding capacity and
system. Chem Rev 89:1765–1800, 1989. brain penetration. Chimia 46:299 –303, 1992.
19. Van de Waterbeemd H, Smith DA, Beaumont K, Walker DK. 43. Abraham MH. Scales of solute hydrogen-bonding: their construc-
Property-based design: optimization of drug absorption and phar- tion and application to physicochemical and biochemical pro-
macokinetics. J Med Chem 44:1313–1333, 2001. cesses. Chem Soc Rev 22:72– 83, 1993.
20. Ekins S. Predicting undesirable drug interactions with promiscuous 44. Palm K, Luthman K, Ungell A-L, Strandlund G, Artursson P.
proteins in silico. Drug Discov Today 9:276 –285, 2004. Correlation of drug absorption with molecular surface properties.
21. Leo A, Hansch C, Elkins D. Partition coefficients and their uses. J Pharm Sci 85:32–39, 1996.
Chem Rev 71:525– 615, 1971. 45. Kelder J, Grootenhuis PDJ, Bayada DM, Delbressine LPC, Ploe-
22. Hansch C, Steward AR, Anderson SM, Bentley D. The parabolic men J-P. Polar molecular surface as a dominating determinant for
dependence of drug action upon lipophilic character as revealed by oral absorption and brain penetration of drugs. Pharm Res 16:
a study of hypnotics. J Med Chem 11:1–11, 1967. 1514 –1519, 1999.
23. Gao H, Hansch C. QSAR of P450 oxidation: on the value of 46. van de Waterbeemd H, Camenish G, Folkers G, Chretien JR,
comparing kcat and Km with kcat/Km. Drug Metab Rev 28:513–526, Raevsky OA. Estimation of blood-brain barrier crossing of drugs
1996. using molecular size and shape, and H-bonding characteristics. J
24. Mertsch K, Maas J. Blood-brain barrier penetration and drug de- Drug Target 6:151–165, 1998.
velopment from an industrial point of view. Curr Med Chem - 47. Clark DE. Rapid calculation of polar molecular surface area and its
Central Nervous System Agents 2:187–201, 2002. application to the prediction of transport phenomena 1. Prediction
25. Darvas F, Dorman G, Urge L, Szabo I, Ronai Z, Sasavari-Szekely of intestinal absorption. J Pharm Sci 88:807– 814, 1999.
M. Combinatorial chemistry. Facing the challenge of chemical 48. Ertl P, Rohde B, Selzer P. Fast calculation of molecular surface
genomics. Pure Appl Chem 73:1487–1498, 2001. area as a sum of fragments-based contributions and its application
26. Strausberg RL, Schreiber SL. From knowing to controlling: a path to prediction of drug transport properties. J Med Chem 42:3714 –
from genomics to drugs using small molecule probes. Science 3717, 2000.
30:294 –295. 49. Österberg T, Norinder U. Prediction of polar surface area and drug
27. Schreiber SL. The small molecule approach to biology. Chem Eng transport processes using simple parameters and PLS statistics.
News, pp 51– 61, 3 March 2003. J Chem Inf Comput Sci 40:1408 –1411, 2000.
28. Ekins S, Waller CL, Swaan PW, Cruciani G, Wrighton SA, Wikel 50. Iyer M, Mishra R, Han Y, Hopfinger AJ. Predicting blood-brain
JH. Progress in predicting human ADME parameters in silico. barrier partitioning of organic molecules using membrane-interac-
J Pharmacol Toxicol Methods 44:251–272, 2000. tion QSAR analysis. Pharm Res 19:1611–1621, 2002.
29. Tute MS. History and objectives of quantitative drug design. In: 51. Veber DF, Johnson SR, Cheng H-Y, Smith BR, Ward KW, Kopple
Comprehensive medicinal chemistry. Vol. 4 (Sammes PG, Taylor KD. Molecular properties that influence the oral bioavailability of
JB, eds.), pp 1–32. Oxford, Pergamon, 1990. drug candidates. J Med Chem 45:2515–2623, 2002.
30. Levin VA. Relationship of octanol/water partition coefficient and 52. Pan D, Iyer M, Liu J, Li Y, Hopfinger AJ. Constructing optimum
molecular weight to rat brain capillary permeability. J Med Chem blood brain barrier QSAR models using a combination of 4D-
23:682– 684, 1980. molecular similarity measures and cluster analysis. J Chem Inf
31. Young RC, Durant GJ, Emmett JC, Ganellin CR, Graham MJ, Comput Sci 44:2083–2098, 2004.
Mitchell RC, Prain H D, Roantree ML. Dipole moment in relation 53. Leeson PD, Davis AM. Tine-related differences in the physical
to hydrogen receptor histamine antagonist activity for cimetidine property profiles of oral drugs. J Med Chem 47:6338 – 6348, 2004.
analogs. J Med Chem 29:44 – 49, 1986. 54. Hansch C, Leo AJ. Substituent constant for correlation analysis in
32. Young RC, Mitchell RC, Brown TH, Ganellin CR, Griffiths R, chemistry and biology. New York: Wiley, 1979.
Jones M, Rana KK, Saunders D, Smith IR, Sore NE, Wilks TJ. 55. Fichert T, Yazdanian M, Proudfoot JR. A structure-permeability
Development of a new physicochemical model for brain penetra- study of small drug-like molecules. Bioorganic Med Chem Lett
tion and its application to the design of centrally acting H2 receptor 13:719 –722, 2003.
antagonists. J Med Chem 31:656 – 671, 1988. 56. Van de waterbeemed, Atkinson F, Cole S, Green C. Lipophilicity
33. ter Laak AM, Tsai RS, Donné-Op den Kelder GM, Carrupt P-A, and other parameters affecting brain penetration. Curr Med Chem-
Testa B, Timmerman H. Lipophilicity and hydrogen-bonding ca- Central Nervous System Agents 2:229 –240, 2002.
pacity of H1-antihistaminic agents in relation to their central sed- 57. Hansch C, Bjorkroth J, Leo AJ. Hydrophobicity and central ner-
ative side-effects. Eur J Pharm Sci 2:373–384, 1994. vous system agents: on the principle of minimal hydrophobicity in
34. Goodwin JT, Conradi RA, Ho NFH, Burton PS. Physicochemical drug design. Pharm Sci 76:663, 1987.
determinants of passive membrane permeability: role of solute 58. Feng RM. Assessment of blood-brain barrier penetration: in silico,
hydrogen-bonding potential and volume. J Med Chem 44:3721– in vitro and in vivo. Curr Drug Metab 3:647– 657, 2002.
3729, 2001. 59. Skaaeda T, Okamura N, Nagata S, Yagami T, Horinouchi M,
35. Taylor PJ. Hydrophobic properties of drugs. In: Comprehensive Okumura K, Yamahita F, Hashida M. Molecular and pharmaco-

NeuroRx威, Vol. 2, No. 4, 2005

 SUCCESSFUL CNS DRUGS 553

kinetic properties of 222 commercially available oral drugs in 71. Redfern SW, Carlsson L, Davis SA, Lynch GW, MacKenzie I,
humans. Biol Pharm Bull 24:935–940, 2001. Palethorpe S, Siegl S KP, Strang I, Sullivan TA, Wallis R. Rela-
60. Abraham MH, Chadha HS, Martins F, Mitchell RC, Bradbury tionships between preclinical cardiac electrophysiology, clinical
MW, Gratton JA. Hydrogen bonding part 46: a review of the QT interval prolongation and torsade de pointes for a broad range
correlation and prediction of transport properties by an LFER of drugs: evidence for a provisional safety margin in drug devel-
method: physicochemical properties brain penetration and skin opment. Cardiovasc Res 58:32– 45, 2003.
permeability. Pestic Sci 55:78 – 88, 1999. 72. Adenot M, Lahana R. Blood-brain barrier permeation models:
61. Clark DE. In silico prediction of blood-brain barrier permeation, discriminating between potential CNS and non-CNS drugs includ-
Drug Discovery Today 8:927–933, 2003. ing P-glycoprotein substrates. J Chem Inf Comp Sci 44:239 –248,
62. Austin RP, Davis AM, Manners CN. Partitioning of ionizing mol- 2004.
ecules between aqueous buffers and phospholipids vesicles. 73. Sadowski J, Kubinyi H. A scoring scheme for discriminating be-
J Pharm Sci 84:1180 –1183, 1995. tween drugs and nondrugs. J Med Chem 41:3325–3329, 1998.
63. Lin JH, Rodrigues AD. In vitro model for early studies of drug
74. Ajay, Walters WP, Murcko MA. Can we learn to distinguish
metabolism. In: Pharmacokinetics optimization in drug research:
between “drug-like” and “nondrug-like” molecules? J Med Chem
biological, physicochemical and computational strategies (Testa,
Van de Waterbeemed H, Folker G, Guy R, eds.), pp 217–243. New 41:3314 –3324, 1998.
York: Wiley-VCH, 2001. 75. Ajay, Walters WP, Murcko MA. Designing libraries with CNS
64. Graaf deC, Vermeulen EPN, Feenstra NK. Cytochrome P450 in activity. J Med Chem 41:4942– 4951, 1998.
silico: an integrative modeling approach. J Med Chem 48:2726 – 76. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experimental
2755, 2005. and computational approaches to estimate solubility and perme-
65. Ertel SI, Ertel EA, Clozel JP. T-type Ca2⫹ channels and pharma- ability in drug discovery and development settings. Adv Drug
cological blockade: potential pathophysiological relevance. Car- Deliv Rev 23:3–25, 1997.
dioavasc Drugs Ther 11:723–739, 1997. 77. Wenlock MC, Austin RP, Barton P, Davis AM, Leeson PD. A
66. Van de waterbeemed H. The fundamental variables of the biophar- comparison of physiochemical property profiles of development
maceutics classification system (BCS): a commentary. Eur and marketed oral drugs. J Med Chem 46:1250 –1256, 2003.
J Pharm Sci 7:1–3, 1998. 78. Lipinski CA. Drug-like properties and the causes of poor solubility
67. Liljebris C J, Larson SD, Ogg D, Palazuk JB, Bleasdale E. Inves- and poor permeability. J Pharmacol Toxicol Methods 44:235–249,
tigation of potential bioisosteric replacements for the carboxyl 2000.
groups of peptidomimetic inhibitors of protein tyrosine phospha- 79. Vieth M, Siegel MG, Higgs RE, Watson IA, Robertson DH, Savin
tase 1B: identification of a tetrazole-containing inhibitor with cel- KA, Durst GL, Hipskind PA. Characteristic physical properties
lular activity. J Med Chem 45:1785–1798, 2002. and structural fragments of marketed oral drugs. J Med Chem
68. Feng RM, Assessment of blood-brain barrier penetration: in silico, 47:224 –232, 2004.
in vitro and in vivo. Curr Drug Metab 3:647– 657, 2002. 80. Lipinski CA. Drew University Medical Chemistry Special Topics
69. Raub JT. Early preclinical evaluation in support of hit identifica- Course. July 1999.
tion and lead optimization for brain exposure. AAPS Workshop on 81. Lipinski CA. Lead- and drug-like compounds: the rule-of-five
Optimization of Drug-Like Properties During Lead Optimization. revolution. Drug Discov Today Technol 1:337–341, 2004.
Parsippany, NJ, 19 –22 September, 2004. 82. Mahar Doan KM, Humphreys JE, Webster LO, Wring SA, Shamp-
70. Recanatini M, Poluzzi E, Masetti M, Cavalli A, De Ponti F. QT ine LJ, Serabit-Singh CJ, Adkinson KK, Polli JW. Passive perme-
prolongation through hERG K⫹ channel blockade: current knowl- ability and P-glycoprotein-mediated efflux differentiate central
edge and strategies for the early prediction during drug develop- nervous system (CNS) and non-CNS marketed drugs. J Pharmacol
ment. Med Res Rev 25:133–166, 2005. Exp Ther 303:1029 –1037, 2002.

NeuroRx威, Vol. 2, No. 4, 2005