Optimization of Lead Structures

A lead structure is the starting point on the way to a drug. The potency, specificity,
and duration of effect must be optimized, and the side effects and toxicity must be
minimized in an usually elaborate, iterative process. Every change in the chemical
structure modulates the 3D structure of the molecule, its physicochemical prop-
erties, and the activity spectrum. The isosteric replacement of atoms or groups,
the introduction of hydrophobic building blocks, the dissection of rings or the
restriction of flexible molecular portions into cyclic structures, and the optimiza-
tion of the substitution pattern are all possibilities to purposefully modify a target
structure.
Creativity and luck are always important prerequisites for success in pharmaceu-
tical research. Nonetheless, there is a treasure chest of decades of accumulated
experience that can be exceedingly supportive to the rational optimization process.
The computer-aided methods can contribute to their full capability in this field in
particular. Several general considerations and approaches to lead optimization are
presented in the sections of this chapter. A discussion of the structure-based
and computer-aided optimization of lead structures is presented in ▶ Chaps. 17,
“Pharmacophore Hypotheses and Molecular Comparisons” and ▶ 20, “Protein
Modeling and Structure-Based Drug Design”; examples for its application to differ-
ent therapeutic areas are presented in ▶ Chaps. 23, “Inhibitors of Hydrolases with an
Acyl–Enzyme Intermediate”; ▶ 24, “Aspartic Protease Inhibitors”; ▶ 25, “Inhibitors
of Hydrolyzing Metalloenzymes”; ▶ 26, “Transferase Inhibitors”; ▶ 27, “Oxidore-
ductase Inhibitors”; ▶ 28, “Agonists and Antagonists of Nuclear Receptors”; ▶ 29,
“Agonists and Antagonists of Membrane-Bound Receptors”; ▶ 30, “Ligands for
Channels, Pores, and Transporters”; ▶ 31, “Ligands for Surface Receptors”; ▶ 32,
“Biologicals: Peptides, Proteins, Nucleotides, and Macrolides as Drugs”.

8.1 Strategies for Drug Optimization

The optimization of active substances follows a process that is best characterized by

the words of the philosopher Sir Karl Popper:

G. Klebe, Drug Design, DOI 10.1007/978-3-642-17907-5_8, 153

# Springer-Verlag Berlin Heidelberg 2013
154 8 Optimization of Lead Structures

The truth is objective and absolute. But we can never be sure that we have found it. Our
knowledge is always an assumed knowledge. Our theories are hypotheses. We test for the
truth in that we exclude what is false. (Objective Knowledge, 1972)

Accordingly the optimization of a compound’s potency follows a working

hypothesis, while an iterative process of trial and error refines the hypothesis. The
assembled data about the relationship between chemical structure and biological
activity serve the design of new structures. These are synthesized and tested, and
a new working hypothesis is modified as appropriate. In negative cases, the
hypothesis is discarded and a new one is formulated that fits more harmoniously
with the biological data. The following qualities in the structure of the active
substance are distinguished from one another:
• The actual pharmacophore (Sects. 8.7 and ▶ 17.1) that is responsible for the
specific binding and upon which only limited chemical modification can be
carried out,
• The additional groups (adhesion groups) that improve the affinity and biolog-
ical activity,
• Further groups that do not influence the binding but rather the lipophilicity of
the molecule and with it the transport and distribution in biological systems
(▶ Chap. 19, “From In Vitro to In Vivo: Optimization of ADME and Toxicology
Properties”),
• The groups that must be cleaved or modified in the organism to release the
actual active form (▶ Chap. 9, “Designing Prodrugs”).
The most important steps in the optimization of lead structures are the systematic
changes in the shape and form, that is, the three-dimensional structure, and/or the
physicochemical properties. Single steps along this route are:
• Changes in the lipophilicity and the electronic properties through the introduc-
tion or removal of hydrophobic or hydrophilic groups,
• Variations of substituents at aromatic or heteroaromatic rings,
• Introduction or elimination of heteroatoms in chains or rings,
• Changes in chain length of aliphatic groups or linkers,
• Introduction of space-filling substituents to stabilize a particular conformation,
• Changes in the ring size of alicyclic or heterocyclic rings,
• Incorporation of flexible partial structures in rings,
• Incorporation of branches or attachments to rings (rigidifying),
• Opening of rings,
• Elimination of chiral centers to simplify a structure,
• Addition of chiral centers to increase the selectivity or
• Shift the thermodynamic binding profile and the drug’s residence time at the
target protein.
These processes are usually unidirectional in classical drug optimization,
that is, the optimization takes place on one position of the molecule at a time, in
one single direction. In the past, such unidirectional optimization has led to many
disappointments because interdependent influences of the structural changes were
neglected, or the optimal lipophilicity was exceeded. John Topliss developed
8.2 Isosteric Replacement of Atoms and Functional Groups 155

a scheme for the variation of aromatic substituents that allows the biological activity
to be optimized in a minimum number of steps (Sect. 8.3). The application of
experimental design, simultaneously changing multiple parts of a molecule, and the
evaluation of the results by using quantitative structure–activity relationships
(▶ Chap. 18, “Quantitative Structure–Activity Relationships”) usually allows a fast
and effective optimization. In structure-based and computer-aided optimization, the
3D structure of the target protein and its complexes leads to directed structural
variations of the active substances. Here again, the aspects of total lipophilicity and
metabolism should not be neglected.

8.2 Isosteric Replacement of Atoms and Functional Groups

Isosteric replacement is the exchange of particular groups in a molecule for

sterically and electronically related groups. If the biological effect is essentially
maintained, the term bioisosteric replacement (Fig. 8.1) is used. In the simplest
case a single atom is exchanged, for instance, a Cl (lipophilic, weakly electron
withdrawing) is replaced by a Br (same characteristics as Cl) or methyl (lipophilic,
weakly electron donating), or an –O– (polar, H-bond acceptor) is exchanged for an
NH (polar, H-bond donor) or a –CH2– (lipophilic, unable to form H-bonds).
Furthermore, bioisosteric replacement also means the exchange of entire groups.

Substituents: F-, Cl-, Br-, CF3-, NO2-

Methyl-, Ethyl-, Isopropyl-, Cyclopropyl-, tert-Butyl-,
-OH, -SH, -NH2, -OMe, -N(Me)2

Bridging Groups: -CH2-, -NH-, -O-

-COCH2-, CONH-, -COO-,
>C=O, >C=S, >C=NH, >C=NOH, >C=NOAlkyl

Atoms and Groups in Rings: -CH=, -N=

-CH2-, -NH-, -O-, -S-
-CH2CH2-, CH2-O- -CH=CH-, -CH=N-

Larger Groups: -NHCOCH3, -SO2CH3 H

N N N O
-COOH, -CONHOH, -SO2NH2, NH , O
N N
HO HO N

HO N N
H H

Fig. 8.1 A few possibilities for the isosteric replacement of atoms and/or groups.
156 8 Optimization of Lead Structures

I I

HO O CH2CH(NH2)COOH

I 8.1 Triiodothyronine, T3

HO O CH2CH(NH2)COOH

8.2
R
COOH
O

O 8.4 R = -COOH
NH2
8.3 Acetylsalicylic acid or -SO2NH2

Fig. 8.2 Isosteric replacement with retention, loss, and reversal of the biological activity. All
three iodine atoms of the thyroid hormone thyroxine 8.1 can be replaced with alkyl groups and
compound 8.2 is still active. In the case of acetylsalicylic acid 8.3, the exchange of the –OCOCH3
for an NHCOCH3 group led to the loss of the acylating ability and therefore a nearly complete loss of
the biological activity. The antimetabolite sulfanilamide 8.4 (R ¼ SO2NH2) is derived from
p-aminobenzoic acid 8.4 (R ¼ COOH), which is a critical intermediate in the bacterial dihydrofolate
synthesis; 8.4 (R ¼ SO2NH2) is the result of the exchange of a carboxyl group for an isosteric
sulfonamide group.

For example, –COOH, an H-bond acceptor and donor, can be replaced with other
groups that have the same or modified properties, for instance, with the similarly
acidic tetrazole. Another example can be found in the exchange of a phenyl ring for
a thiophene or a furan building block (Fig. 8.1). The potential of isosteric replace-
ment is illustrated in the exchange of all three iodine atoms of triiodothyronine T3
8.1 for alkyl groups to give 3,5-dimethyl-30 -isopropylthyronine 8.2, which in turn
retains impressive affinity and agonistic activity on the thyroid hormone receptor.
In contrast to triiodothyronine, which is both iodinated and metabolized by
a deiodinase, the alkyl groups of 8.2 are no longer metabolically cleavable.
Bioisosteric replacement was and is one of the most important strategies in
pharmaceutical research. Nonetheless, surprises sometimes occur. The replacement
of an ester for an amide group in the local anesthetics (▶ Sect. 3.4) expectedly
improved the metabolic stability. In the case of acetylsalicylic acid 8.3 (Fig. 8.2)
this exchange cannot be made. An analogous exchange of the –COO– group for
a –CONH– group results in a complete activity loss because the amide can no
longer acylate the cyclooxygenase enzyme (▶ Sect. 27.9). In the case of
p-aminobenzoic acid (R ¼ –COOH, Fig. 8.2) the exchange of a carboxyl group
for a sulfonamide group gives sulfanilamide 8.4 (R ¼ –SO2NH2), which is an
antimetabolite of p-aminobenzoic acid (▶ Sect. 2.3).
8.3 Systematic Variation of Aromatic Substituents 157

A lead structure is rarely studied exclusively by one research group. Other

companies adopt successful examples, at the very latest after the economic success
of a new medicine. The goal of this so-called “me-too” research is to modify the
competitor’s lead structure to arrive at patent-free analogues that are more effica-
cious, more selective, or better tolerated. It must be accepted that even this form of
competition has led to the therapeutically most valuable compounds in many thera-
peutic areas. On the one hand, a plentitude of duplicate work has been performed,
while on the other hand, new analogues with improved properties have been produced
and introduced to therapy which turned out to be successful in the long run. Penicil-
lins of the third and fourth generation with broad-spectrum activity and metabolic
stability, b-blockers with improved selectivity, and many other specific drugs would
simply not exist if it were not for the much-disparaged “me-too” research.

8.3 Systematic Variation of Aromatic Substituents

The goal of lead structure optimization has an impact on the planning of the
relevant experimental series. If the biological consequences of structural changes
are to be evaluated with minimal effort, careful design must precede the synthesis
of the substances. Here an almost unsolvable problem emerges in that, as a general
rule, the exchange of a substituent or group leads to complex changes in multiple
properties. The exchange of an ethyl group for a methyl group changes only the
lipophilicity and size of the substituent. If a methyl group is exchanged for
a chlorine atom, the polarizability, electronic properties, and moreover the metab-
olism is altered. Other substituents could then change the H-bond donor and
acceptor properties as well as the ionization and dissociation.
In 1971, Paul Craig proposed the use of a simple diagram for the structural variation
of aromatic substituents, with which the important characteristics of these substitu-
ents, for instance, lipophilicity and electronic properties, are plotted against each
other. The selection of substituents from different quadrants of this diagram allows
an evaluation of different combinations of properties. The concept can be extended to
multiple dimensions, possibly with the aid of mathematical and statistical methods.
In 1972, John Topliss made a suggestion that went further, which would be
called today an evolutionary strategy. One substituent at a time (e.g., hydrogen for
chlorine) is exchanged in the optimization of the substitution pattern of an aromatic
compound. The next compound is planned based on which of the first two com-
pounds demonstrated better effects. If the new substituent improves the effect,
a new substituent is chosen that has the same physicochemical properties, in larger
measure, or more of these substituents are added. If the new substituents impair the
biological activity, then a substituent is chosen that has the opposite physicochem-
ical properties. If two different substituents produce the same effect, it should be
evaluated whether changes in the physicochemical properties influence the activity
in the opposite direction. Despite its elegance, this strategy often fails for the
mundane reason that it is too time consuming to take such a stepwise approach.
158 8 Optimization of Lead Structures

As a consequence of the work of Craig and Topliss, further design methods were
developed. None of these methods should be interpreted too closely. Synthetic
planning must be oriented on both the accessibility of the compounds as well as
achieving the largest possible structural variation, that is, a diversity of physico-
chemical properties and 3D structure. Since the introduction of combinatorial
chemistry (▶ Chap. 11, “Combinatorics: Chemistry with Big Numbers”), the ratio-
nal design of diverse substance libraries has taken on entirely new possibilities and
perspectives.

8.4 Optimizing the Activity and Selectivity Profile

The structural variation of a lead structure influences not only the activity strength
but also the activity spectrum. That can be thoroughly advantageous, but it also
brings with it the risk that the selectivity can deteriorate. A simple rule of thumb is
that enlarging the molecule, introducing optically active centers, and rigidification
improves the selectivity, assuming that the activity is not entirely lost. On the other
hand, removing a chiral center, establishing more flexibility, or reducing the size of
the molecule usually results in unspecific and weaker activity.
Because of the sequencing of the human genome, the gene family to which
a target protein belongs is known, as is the number of members of the gene family.
By using gene technology it is possible to construct single isoform test systems
(assays). As a result, today pharmaceutical research is in a position to make
a predictive selectivity profile. This has stimulated efforts to develop selective
drugs. An interesting corollary to these efforts is the fact that the molecular weight
of drugs has increased, as statistics show, in the last years, a confirmation of the
above-mentioned rule of thumb.
For drugs that are meant to act on neuroreceptors in the brain, the polarity is critical
to whether they can cross the blood–brain barrier. Polar compounds are unable to do
this and act only in the periphery, for instance, on the circulatory system. Examples of
this are adrenaline 8.5 and dopamine 8.6 (Fig. 8.3). The stepwise removal or masking
of polar groups brings the central effects into the foreground. Ephedrine 8.7 acts in the
brain and in the periphery, it is centrally stimulating and raises the blood pressure.
Amphetamine 8.8 (“speed”) and the intoxicant MDMA 8.9 (the designer drug
“ecstasy”) are weak bases. Their relatively nonpolar neutral forms easily overcome
the blood–brain barrier and their CNS effects dominate (Fig. 8.3).
There are exceptions even here. L-DOPA 8.10 (Fig. 8.3) is an extremely polar
amino acid. It could never cross the blood–brain barrier by passive diffusion alone.
Instead it is recognized by an amino acid transporter and actively transported over
the membrane and into the brain. This simultaneously solves the problem of
bringing dopamine 8.6, which is used to treat Parkinson’s disease, into the brain
because L-DOPA is decarboxylated to dopamine there (▶ Sects. 9.4 and ▶ 27.8).
The decisive influence that even the smallest changes in the structure can have is
seen in the effect spectrum of the hormone and neurotransmitter noradrenaline and
adrenaline and their synthetic analogues. Whereas noradrenaline 8.11 (Fig. 8.4)
8.4 Optimizing the Activity and Selectivity Profile 159

Polar Molecules Intermediate Polarity: Nonpolar Molecules:

OH
H NH2
HO N
CH3 CH3
OH
HO H
N 8.8 Amphetamine
CH3
8.5 Adrenaline
CH3
H
HO NH2 8.7 Ephedrine N
O CH3
R CH3
HO O

8.6 Dopamine, R = H 8.9 MDMA

8.10 L-DOPA, R = COOH

Fig. 8.3 The polar compounds adrenaline 8.5 and dopamine 8.6 are cardiovascularly active in the
periphery after intravenous administration. Ephedrine 8.7 is more lipophilic and therefore shows
both peripheral and central effects. The more nonpolar compound amphetamine 8.8 (“speed”) has
overwhelmingly stimulatory effect in the CNS. 3,4-Methylenedioxymethamphetamine 8.9
(MDMA; “ecstasy”) is hallucinogenic. Polar groups are red and neutral or lipophilic groups are
blue.

8.11 Noradrenaline, R = H
OH
H Predominantly α-Mimetic
HO N 8.5 Adrenaline, R = CH3
R
α- and β-Mimetic
HO
8.12 Isoprenaline, R = -CH(CH3)2
β1-Mimetic

OH
H 8.13 Dobutamine
HO N
β1-Mimetic
CH3
HO
OH OH
H H
N CH3 Cl N CH3
HO
CH3 CH3
CH3 CH3
HO H2N
Cl

8.14 Salbutamol 8.15 Clenbuterol

b2-Mimetic b2-Mimetic

Fig. 8.4 Noradrenaline 8.11, adrenaline 8.5, and isoprenaline 8.12 act to different extents on the
a and b receptors. Selective b1 and b2 agonists, for instance, 8.13, 8.14, and 8.15, act specifically as
cardiac stimulants or bronchodilators.
160 8 Optimization of Lead Structures

O O O O O O O
S S S
H2N NH H2N OH

N O
Cl N Cl
H H

8.16 Hydrochlorothiazide 8.17 Furosemide

O O O
S 8.18 Carbutamide, R = NH2
N N CH3
H H 8.19 Tolbutamide, R = CH3
R

O O O
S
O N N
H H
Cl
N
H
OMe 8.20 Glibenclamide

Fig. 8.5 The sulfonamides hydrochlorothiazide 8.16, furosemide 8.17, and related diuretics are
different from most antibacterial analogues because of the unsubstituted sulfonamide group.
Carbutamide 8.18 and tolbutamide 8.19 were the first unspecific sulfonamides with hypoglycemic
effects that were later replaced with specific hypoglycemics of the glibenclamide-type 8.20.

affects the a-adrenergic receptors, its N-methyl derivative adrenaline 8.5 (Fig. 8.3)
acts on a and b receptors as a mixed a/b agonist. This difference was used to
enlarge the N-alkyl group to arrive at the specific b-agonist isoprenaline 8.2
(Fig. 8.4). Further differentiation of the effects could be achieved within the class
of b-adrenergic substances. Dobutamine 8.13 is missing the alcoholic hydroxyl
group of adrenaline. Despite its structural relationship to dopamine 8.6 (Fig. 8.3) it
is a b1 agonist with cardioselective effects. Specific b2 agonists, for instance
salbutamol 8.14 and clenbuterol 8.15 (Fig. 8.4) are used to treat asthma because
they are bronchiodilators without the cardio-stimulatory effects of the unspecific b
agonists (▶ Sect. 29.3).
The sulfonamides are a prime example for the targeted optimization of lead
structures in different therapeutic indications. From the first antibacterial examples,
the diuretics as well as hypoglycemics (antidiabetics) resulted. It had already been
noticed in 1940 that sulfanilamide (▶ Sect. 2.3) inhibits the enzyme carbonic
anhydrase, and therefore should lead to increased urine production (▶ Sect. 25.7).
Among other substances, hydrochlorothiazide 8.16, furosemide 8.17 (Fig. 8.5), and
structurally related compounds gained therapeutic importance. In the early 1940s,
the hypoglycemic effects of a few sulfonamides were clinically observed. The
antibacterial and simultaneously hypoglycemic carbutamide 8.18 was introduced
to therapy in 1955, the lipophilic and therefore more bioavailable tolbutamide 8.19
8.5 From Agonists to Antagonists 161

was introduced later. Systematic structural variation finally led to glibenclamide

8.20 (Fig. 8.5 and ▶ Sect. 30.2), which is much more potent and specific.

8.5 From Agonists to Antagonists

There is no general recipe for the transformation of an agonist into an antagonist. An

example of this is found in the tedious route from the agonist histamine to the H2
antagonist, as is described in detail in ▶ Sect. 3.5. There are, however, recognized
principles that have proven to be of value. For example, the exchange of polar for non-
polar substituents or the introduction of large groups such as additional aromatic rings
changes some receptor agonists to antagonists. The exchange of both phenolic
hydroxyl groups in isoprenaline 8.12 for two chlorine atoms (DCI, 8.21) or additional
aromatic rings (pronethalol, 8.22) delivered the first b-adrenergic antagonists, the
b-blockers. The introduction of an oxygen atom in the side chain, and further structural
optimization afforded the first b1-selective antagonists, for example, practolol 8.23 and
metoprolol 8.24. The b1-selective partial agonist xamoterol 8.25 is a blocker as well as
an agonist (Fig. 8.6). It occupies b1 receptors and displays a moderately stimulating
effect. By occupying the receptor, it protects it from an excessive response upon
elevated adrenaline release, for instance, from exercise or stress.
Analogously, the exchange of the imidazole ring of histamine 8.26 for large
hydrophobic groups led to the first H1 antagonists, for instance, diphenhydramine
8.27 (Fig. 8.7). Sedation is the most troublesome side effect of the classic H1
antagonists, which are used to treat allergies. The non-sedating terfenadine

OH OH
H H
Cl N CH3 N CH3

CH3 CH3
Cl
8.21 DCI 8.22 Pronethalol

OH
H
O N CH3 8.23 Practolol, R = -NHCOCH3

CH3 8.24 Metoprolol, R = -CH2CH2OMe

R

OH O
H
O N
N N 8.25 Xamoterol
H
O
HO

Fig. 8.6 3,4-Dichloroisoprenaline 8.21 (DCI) and pronethalol 8.22, the first unspecific
b-blockers, were derived from isoprenaline 8.12. Practolol 8.23 and metoprolol 8.24 are specific
b1 agonists. Xamoterol 8.25 is a partial b1 agonist, a combined agonist and antagonist.
162 8 Optimization of Lead Structures

Fig. 8.7 By starting with

H
histamine 8.26 and N NH2 O CH3
introducing large N
hydrophobic groups, the H1 N
antagonists, for instance, CH3
diphenhydramine 8.27, were 8.26 Histamine
obtained. The non-sedating H-Agonist
terfenadine 8.28 (R ¼ CH3)
crosses the blood–brain 8.27 Diphenhydramine
barrier but is immediately Non-polar H1 Antagonist
expelled by a transporter. (sedating)
In the meantime the active
metabolite, fexofenadine with
OH
R ¼ COOH, is in the market.
N

OH R
CH3
CH3
8.28 Terfenadine, T = CH3
Polar H1 Antagonist (non-sedating)
Fexofenadine, Active Metabolite: R = -COOH

8.28 (R ¼ H) can cross the blood–brain barrier because of its high lipophilicity, but
is immediately expelled by a transporter. Because of its cardiotoxicity, terfenadine
has been withdrawn from the market in the meantime and replaced by its active
metabolite fexofenadine 8.28 (R ¼ COOH). The sedating side effects of antihista-
mines also led to neuroleptics and antidepressants (▶ Sect. 1.6). Here, however, the
limits of rational drug optimization are apparent. Promethazine 8.29 is an antihis-
tamine with antiallergic action and sedating side effects. The neuroleptic chlor-
promazine 8.30 is a central depressant and therefore an antipsychotic; the
extraordinarily similar structure of imipramine 8.31 acts, on the other hand, as
a stimulant and is an antidepressant (Fig. 8.8). All three substances have different
mechanisms of action. The introduction of additional aromatic rings to other
receptor agonists, for instance, to the neurotransmitters acetylcholine and dopa-
mine, has led to antagonists (Fig. 8.9).

8.6 Optimizing Bioavailability and Duration of Effect

The absorption of the majority of pharmaceuticals depends only on their

lipophilicity. The more polar the drug, the more poorly it can penetrate the lipid
membrane, and the lower the absorption (▶ Sect. 19.6). Increasing the lipophilicity
improves the absorption (▶ Sect. 19.6). Extremely lipophilic compounds are insol-
uble in water, and the absorption is too slow. Lipophilic acids and bases offer
advantages here, if their acidity constant is not too far away from the neutral point,
pH 7. In their ionized form they are highly water soluble, while in their neutral form,
with which they are in equilibrium, they are lipophilic and membrane penetrable.
8.6 Optimizing Bioavailability and Duration of Effect 163

S S

N N Cl N
CH3

N CH3 CH3
H3C CH3 N N
CH3 CH3

8.29 Promethazine 8.30 Chlorpromazine 8.31 Imipramine

H1 Antagonist Neuroleptic Antidepressant

Fig. 8.8 Closely related structures of active substances can have very different qualitative
activity. Chlorpromazine 8.30, a dopamine antagonist with neuroleptic activity, and imipramine
8.31, a dopamine transporter inhibitor with antidepressant activity, are both derived from
promethazine 8.29, an H1 antagonist with antiallergic activity.

H + D
N NH3 P
Fig. 8.9 The active
substance histamine 8.26 and N A
pharmacophores that are
attributed to it (A acceptor, D 8.26 Histamine Pharmacophore
donor, P positively charged (Positively charged
group). form at pH = 7)

These correlations are discussed in detail in ▶ Sect. 19.5. The molecular size influ-
ences the bioavailability insofar that substances with a molecular weight above
500–600 Da are captured by the liver on the sole grounds of the molecular size, and
are quickly excreted with the bile. Aside from this there are substances that penetrate
the membrane regardless of their polarity. These are taken up into the cell or are
eliminated from the cell by transporters (▶ Sect. 30.7). Among these are structural
analogues of amino acids and nucleosides. Classical strategies to extend the duration
of action are the conversion of free hydroxyl groups to ethers (see ▶ Sect. 9.2), the
replacement of esters with amides, and the replacement of metabolically labile amide
groups with isosteres. In a few cases, such structural changes are associated with
a reduction in potency, which is more than compensated for by a longer duration of
action. In the case of peptides the replacement of L-amino acids with D-amino acids,
the inversion of amide groups, and the replacement of larger structural elements with
peptidomimetic groups (▶ Sect. 10.4) have all proven successful.
The metabolism of aliphatic amino groups can be suppressed with alkyl substi-
tution or branching at the a carbon. Secondary alcohols can be converted to the
more bioavailable tertiary alcohols by introducing an ethinyl group at the same
carbon atom (▶ Sect. 28.5). The introduction of an isosteric fluorine atom in the
para position as a replacement for hydrogen prevents hydroxylation in this position.
If steric considerations do not play a role, the para position can also be blocked
164 8 Optimization of Lead Structures

with a larger group, such as a chlorine atom or a methoxy group. In the hydroxylated
3- and 4-position of the neurotransmitters dopamine, adrenaline, and noradrenaline,
the conversion to the monohydroxylated analogues, 3,5-dihydroxy compounds or to
the NH-isosteric indole group (Fig. 8.1, Sect. 8.2) led to metabolically more stable
and therefore longer-acting compounds.

8.7 Variations of the Spatial Pharmacophore

Rational design is characterized by the fact that the common feature of all
active compounds, and the differences to less potent or inactive analogues
can be derived from the structure of the pharmacophore. A pharmacophore
(Sect. 8.9) is defined as a special arrangement of particular functionalities that
are common to more than one drug and form the basis of the biological activity
(▶ Sect. 17.1).
During the course of rational optimization the molecular scaffold and the sub-
stituents at a pharmacophore are changed to maintain the principle function while
arriving at higher potency or better selectivity. Many computer methods have been
developed to generate ideas for the spatial isomorphic replacement of ligand scaf-
folds. By considering the conformational aspects of the molecules (▶ Chap. 16,
“Conformational Analysis”), they scan databases to find possible candidates that,
despite a different parent scaffold, can place the side chains and interacting groups
in the same spatial orientation. Examples of such approaches are presented in
▶ Sect. 10.8 and ▶ Chap. 17, “Pharmacophore Hypotheses and Molecular Com-
parisons”. But an indirect approach using the protein structure has also been tried.
For this, the spatial structure of the protein–ligand complex is the starting point
from which a part of the binding pocket is cut out, and new building blocks for the
ligand are sought. Subsequently the form and interaction properties of the cut-out
pocket are compared with a database of all known protein–ligand complexes
(▶ Sect. 20.4). If a subpocket is discovered that has similarities to the sought-
after pocket, then ligands that bind there provide an interesting design hypothesis.
The structure of the building blocks that occupy the newly discovered pocket can
generate ideas for isosteric structural elements in a modified ligand.
A different strategy that also considers the pharmacophore can be successful.
In this approach the pharmacophore is retained and only those groups are modified
that affect the pharmacokinetic properties, that is, the transport, distribution,
metabolism, and excretion of a molecule. An efficient and pragmatic strategy is
important. For this, it is essential that not too many changes are made at the same
time, and the changes should not be too biased. With little synthetic effort, a broad
spectrum of physicochemical properties and spatial arrangements should be
covered.
In the meantime it has been established that binding to human plasma proteins
such as serum albumin and the acidic k1-glycoprotein is of decisive importance for
the transport and pharmacokinetic properties of a drug. Therefore binding to these
proteins is considered even in the early phase of drug development (▶ Chap. 19,
8.8 Optimizing Affinity, Enthalpy, and Entropy of Binding and Binding Kinetics 165

“From In Vitro to In Vivo: Optimization of ADME and Toxicology Properties”). On

the other hand, binding to the hERG ion channels (so-called “antitarget”) is
avoided because blocking these channels can lead to arrhythmias (▶ Sect. 30.3).
Drug metabolism is in itself a very important theme and must be considered in
earlier phases of development. The cytochrome P450 enzymes are responsible
for the vast majority of chemical transformations that occur on xenobiotics
(▶ Sect. 27.6). To be able to predict the behavior of drug candidates at this stage
of the development process, the expected interactions with these metabolic
enzymes are evaluated in an early phase of optimization. The expression of P450
enzymes can also be induced by xenobiotics. The trigger for this could be the
binding to a transcription factor like the PXR receptor (▶ Sect. 28.7). Drug candi-
dates binding to this transcription factor can be evaluated early in their development
to avoid this undesirable enhanced metabolism.

8.8 Optimizing Affinity, Enthalpy, and Entropy of Binding

and Binding Kinetics

Generally, the binding affinity to a target protein is primarily improved during

the course of optimization. If multiple candidates are available, the ligand
efficiency (▶ Sect. 7.1) in addition to the chemical accessibility leads the way.
Small, potent lead structures offer legitimate hope that they can be well optimized.
Very small compounds that have nanomolar affinity, despite their low molecular
weight, can be problematic. Most of the time an optimal interaction pattern
is already established. It is then almost impossible to transfer this pattern to another
molecular scaffold. Medicinal chemists have established a set of rules based
on experience (▶ Sect. 4.10). According to these rules it is possible to judge
how much a particular group, if correctly placed, can contribute to the binding
affinity.
It was shown in ▶ Sect. 4.10 that the affinity is a combination of the enthalpic
and entropic contributions. Usually one begins with a lead structure that has
a binding affinity in the micromolar range. Expressed as the Gibb’s free
energy DG, this is usually about 30 kJ/mol. An increase in the binding affinity of
4–5 orders of magnitude causes an improvement in DG of 20–30 kJ/mol. Where
should the screw be turned to optimize a lead structure? Does it make more sense to
improve the binding enthalpy, or is one better advised to improve the binding
entropy? Given the enthalpy/entropy compensation described in ▶ Sect. 4.10, is it
even possible to attempt optimization of both values independently? The prereq-
uisite for using such a concept in the optimization is the determination of both
values of a lead structure. Does this help in the choice of the right candidate for
optimization? In the case that the thermodynamic binding profiles of multiple
alternative lead candidates are known, should enthalpically or entropically
driven binders be chosen for optimization? It is very interesting to compare the
thermodynamic signatures of multiple generations of marketed products. The
binding profiles for HIV protease inhibitors (▶ Sect. 24.3) and HMG-CoA
166 8 Optimization of Lead Structures

kcal/mol ΔG ΔH −TΔS
5

−5

−10

−15

−20
r

ir

ir

ir

ir

ir

ir
vi

vi

vi
av

av

av

av

av

av
na

na

na
in

fin

en

an

an

un
di

ito

pi
qu

el

pr

az

pr

ar
In

Lo
R
N
Sa

Am

Ti

D
At
kcal/mol
5

−5

−10

−15

−20
tin

tin

tin

in

tin
at
ta

ta

ta

ta
st
as

as

as

as
va
uv

av

iv

uv
or
er
Fl

Pr

os
At
C

Fig. 8.10 Between 1995 and 2006, the profile of multiple development generations of HIV
protease inhibitors (upper, for formulae see ▶ Fig. 24.15) and statins as HMG-CoA inhibitors
(lower, for formulae see ▶ Fig. 27.13) could be optimized for their thermodynamic signatures, that
is, the extent to which they are driven by entropy or enthalpy. The free energy DG is shown in red,
the enthalpy DH in blue, and the entropic contribution TDS in green. The more negative the
column becomes, the stronger the binding affinity and the more the profile is determined by
enthalpy or entropy. The initially developed compound such as indinavir, saquinavir, nelfinavir,
and pravastatin were entropic binders; in contrast, the newer derivatives such as darunavir or
rosuvastatin have an improved enthalpic profile.

inhibitors (▶ Sect. 27.3) are displayed in Fig. 8.10. Notably, it has been successful
to shift the profile from initially strongly entropically driven binders to
enthalpically driven ones. This observation suggests that it is initially simpler to
optimize a substance’s entropic binding contribution than its enthalpic contribu-
tion. Most of the time this can be seen in the first lead structure upon which an
enlargement of the hydrophobic surface area leads to better binding. The affinity
that is gained is explained by the displacement of ordered water molecules
(▶ Sect. 4.6). Such contributions are assumed to be entropically favorable.
A strategy of introducing rigid rings can also be pursued. In doing so, the com-
pound loses degrees of freedom. If the geometry of the bound state is correctly
frozen, the binding is improved for entropic reasons. An example of this is the
8.8 Optimizing Affinity, Enthalpy, and Entropy of Binding and Binding Kinetics 167

O O

O CH3 O N
H H3C O
H
N
N N H3C S N
H
CH3 O O
H CO2H
O

8.32 HN NH2
8.33
NH2
HN

ΔG : −42.3 kJ/mol ΔG : −49.2 kJ/mol

ΔH: −6.2 kJ/mol ΔH: −48.5 kJ/mol
−TΔS: −36.1 kJ/mol −TΔS : −0.7 kJ/mol

Fig. 8.11 The rigid thrombin inhibitor 8.32 only has a small number of rotatable bonds. It has an
optimal shape complementarity to the binding pocket of thrombin. Its binding is, for the most part,
entropically driven. On the other hand, the considerably more flexible ligand 8.33 has a higher
enthalpic binding contribution.

binding of the largely rigid thrombin inhibitor 8.32, which binds in an almost
exclusively entropically driven manner to the protein (Fig. 8.11). In contrast, the
decidedly more flexible ligand 8.33 displays a large enthalpic binding contribu-
tion. Compound 8.32 represents the result of an optimization that led to a substance
with single-digit nanomolar binding and an optimal shape complementarity for the
binding pocket of thrombin.
As it seems, in general there are applicable concepts for the entropy-driven optimi-
zation. If one can “always win entropically,” then for theoretical reasons enthalpically
favored lead structures should be preferred as a starting point for optimization.
However, caution is called for here. Why a ligand has a particular thermody-
namic profile must be clarified. The inhibitors 8.34 and 8.35 were discovered in
a virtual screening as aldose reductase inhibitors (Fig. 8.12). The chemical struc-
tures of both ligands are very similar. Nevertheless one is an enthalpically driven
binder, and the other is an entropically driven binder. The crystal structure of both
ligands with the protein delivered the reason: the enthalpically preferred inhibitor
8.34 entraps a water molecule, which mediates binding between the ligand and the
protein, whereas the other one does not. The incorporation of a water molecule is
entropically disfavored, and therefore the profile appears to be that of an enthalpic
binder. A resistance profile for inhibitors against mutants of the viral HIV protease
was investigated in the research group of Ernesto Freire at The Johns Hopkins
University in Baltimore (▶ Sect. 24.5). Interestingly, the result was that resistance
to the entropically favored inhibitors could be developed much faster than to
inhibitors with enthalpic advantages. This observation indicates that it is worth-
while to concentrate on enthalpically favored binders in cases in which resistance
can be expected to develop. In the investigated example the enthalpically driven
168 8 Optimization of Lead Structures

OH OH
O S
N
O O
N
O O2N O N N
O2N
8.34 8.35
ΔG: −35.4 kJ/mol ΔG: −31.3 kJ/mol
ΔH : −25.6 kJ/mol ΔH: −8.7 kJ/mol
−TΔS: −9.8 kJ/mol −TΔS: −22.6 kJ/mol

Fig. 8.12 Compounds 8.34 and 8.35 were discovered in a virtual screening as lead structure for
the inhibition of aldose reductase. Although they are structurally similar, 8.34 is a stronger
enthalpic binder and 8.35 is an entropic binder. The subsequent crystal structure analysis of the
complex with the reductase showed that 8.34 traps a water molecule upon binding, whereas this
was not observed with 8.35. Because the entrapment of a water molecule is entropically unfavor-
able, the binding of 8.34 is enthalpically preferred.

binder 8.33 had a less-rigid scaffold (Fig. 8.11). This allows it to more easily elude
changes that are caused by mutations. It is much more difficult for rigid ligands that
bind for entropic reasons to adapt to such steric modifications.
On the other hand, entropic binders can also have an advantage in escaping
resistance. If a ligand is entropically favored because it adopts multiple binding
modes, and even exhibits large residual mobility in the binding pocket when bound,
this can prove to be beneficial! If the protein tries to change the shape of its binding
pocket through resistance mutations to this inhibitor, an incorporated ligand that is
able to adopt multiple binding modes is left with alternative orientations, which,
despite the mutation, still offer good binding.
If it is clear that a lead structure is an enthalpically driven binder, and
superimposed effects such as the entrapment of water molecules have not distorted
the profile, how is the binding of an enthalpically driven binder optimized? Let us
remember the consideration in ▶ Sects. 4.5 and ▶ 4.8: hydrogen bonds, electrostatic
interactions, and van der Waals contacts determine the binding enthalpy. However,
a change in such an interaction property of a molecule is often coupled with
a compensation of enthalpy and entropy. The result is that DG and the binding
affinity do not change at all! The optimization process can be compared to the act of
getting around the inherent enthalpy/entropy compensation. Enthalpically favorable
hydrogen bonds should have an optimal geometry and should not induce severe
structural changes in the protein environment. Otherwise this can lead to an entropic
compensation by causing a shift in the dynamic degrees of freedom. It seems to be
more favorable to strengthen the hydrogen bonds in structurally rigid regions of the
binding pocket. There, enthalpy is better gained because the compensatory shift in
dynamic parameters is less likely. Introduced hydrogen bonds should also not reduce
the degree of desolvation of a bound ligand in that they induce small structural
changes in the binding geometry of hydrophobic groups that become stronger when
8.9 Synopsis 169

exposed to the surrounding solvent environment. It is also important that the local
water structure in the binding pocket remains unchanged.
Another essential question has to do with the optimal interaction kinetics that
a ligand should have. Surface plasmon resonance was introduced in ▶ Sect. 7.7.
The question of whether a ligand binds quickly or slowly to a protein and with what
rate it is released again can be determined with this method. Ideally, how long
should a ligand stay bound to a protein, what is the optimal residence time? The
binding affinity is determined by the relative ratio of the association rate (kon) and
the dissociation rate (koff). It has been shown that structurally similar ligands can
have entirely different kinetic profiles. Which profile is optimal? A loss in affinity
can manifest itself as an increased dissociation rate, or a slower association rate, as
well as a combination of both effects. It was shown in the research group of Helena
Danielson in Uppsala that different binding profiles of therapeutically used HIV
protease inhibitors correlate with the development of resistance to mutants of the
protease. They also demonstrated that resistance forms more rapidly against drugs
that have a higher dissociation rate. This is a decisive criterion to direct drug
optimization in the correct direction. Certainly the kinetic binding profile must be
granted a greater priority in the future. Therefore, a more comprehensive correla-
tion between the structure and the binding is necessary so that this knowledge can
be used for targeted design. Until now, what differentiates a “fast” or “slow” binder
has only been understood in a very few cases. These are parameters that have to do
with the induced-fit adaptations of the protein. It can also involve the ease with
which the desolvation of the previously uncomplexed binding pockets takes place
or with the kinetics with which a ligand in the solvated state sheds its own water
shell. More attention must be paid to these protein and ligand-based properties.

8.9 Synopsis

• A lead structure is only the starting point on the way to a drug; potency,
specificity, and duration of action have to be optimized concurrently to minimize
side effects and toxicity.
• The structure of an active substance is determined by its pharmacophore, which
is responsible for target binding. Its adhesion groups enhance potency and
biological activity, its lipophilicity is responsible for transport and distribution,
and groups to be cleaved or modified release the active form.
• Multiple concepts to modify the chemical structure of a lead can be planned,
however, optimization is multifactorial due to highly correlated influences of the
attempted changes.
• Bioisosteric functional group replacement attempts the exchange of groups on
a given skeleton for sterically and electronically related groups that maintain
activity but improve other drug properties.
• Me-too research follows the goal of modifying the competitor’s lead structures
to arrive at patent-free analogues with improved properties.
170 8 Optimization of Lead Structures

• Assuming unchanged activity, enlarging a molecule, adding chiral centers,

and rigidification usually improves selectivity, whereas removing chiral
centers, allowing more flexibility, and reducing the size makes a drug less
selective.
• The activity spectrum of a substance can be tailored even by the smallest
structural changes that modulate affinity, transportation, distribution, or metab-
olism. Therefore a particular compound class can show activity in quite different
therapeutic indications.
• Transforming agonists to antagonists does not follow clear-cut rules, however,
increasing the size and the attachment of hydrophobic groups such as aromatic
rings often shift the profile.
• The more polar a drug, the more poorly it can penetrate lipid membranes, and the
lower is the absorption. On the other hand, special transporters can assist
penetration.
• Extension of the duration of action is mostly achieved by replacement of
metabolically labile groups with more stable isosteres, the introduction of
more branching groups, blockage of metabolically labile positions at aromatic
rings by F or Cl, or by exchanging L- for D-amino acids concurrently with the
inversion of amide groups.
• Molecular databases can be screened to detect other scaffolds or substitution
patterns that represent a given pharmacophore in an alternative fashion.
• In the early phase of drug development undesired binding to plasma proteins,
antitargets such as the hERG ion channel or preferred binding, inhibition, or
activation of transcription factors or metabolizing cytochrome P450 enzymes are
examined and possibly avoided.
• Proper adjustment of the thermodynamic binding profile can be essential for the
optimization of binding affinity and to endow a drug with the required target-
specific properties. Similarly the interaction kinetics determining binding on and
off rates or residence times are of decisive importance to develop drugs with, for
example, an optimal resistance profile.

Bibliography

General Literature
Sneader W (1985) Drug discovery: the evolution of modern medicines. Wiley, New York
Taylor JB, Triggle DJ (eds) (2007) Comprehensive medicinal chemistry II. Elsevier, Oxford
Wermuth CG (ed) (2008) The practice of medicinal chemistry, 3rd edn. Elsevier-Academic,
New York

Special Literature

Burger A (1991) Isosterism and bioisosterism in drug design. Fortschr Arzneimittelforsch

37:287–371
Bibliography 171

Copeland RA, Pompliano DL, Meek TD (2006) Drug–target residence time and its implications
for lead optimization. Nat Rev Drug Discov 5:730–740
Fokkens J, Klebe G (2006) A simple protocol to estimate protein binding affinity differences for
enantiomers without prior resolution of racemates. Angew Int Ed Engl 45:985–989
Hansch C (1974) Bioisosterism. Intra-Science Chem Rept 8:17–25
Lipinski CA (1986) Bioisosterism in drug design. Ann Rep Med Chem 21:283–291
Ohtaka H, Freire E (2005) Adaptive inhibitors of the HIV-1 protease. Prog Biophys Mol Biol
88:193–208
Shuman CF, Markgren P-O, H€am€al€ainen M, Danielson UH (2003) Elucidation of HIV-1 protease
resistance by characterization of interaction kinetics between inhibitors and enzyme variants.
Antiviral Res 58:235–242
Steuber H, Heine A, Klebe G (2007) Structural and thermodynamic study on aldose reductase:
nitro-substituted inhibitors with strong enthalpic binding contribution. J Mol Biol 368:618–638
Thornber CW (1979) Isosterism and molecular modification in drug design. Chem Soc Rev
8:563–580