Statistical molecular design, parallel synthesis, and biological evaluation of a library of thrombin inhibitors

3424

J . Med. Chem. 2001, 44, 3424-3439

Sta tistica l Molecu la r Design , P a r a llel Syn th esis, a n d Biologica l Eva lu a tion of a

Libr a r y of Th r om bin In h ibitor s

Anna Linusson,*^,†J ohan Gottfries,^†Thomas Olsson,^†Eivor O¨ rnskov,^†Staffan Folestad,^†Bo Norde´n,^‡and

Svante Wold^§

AstraZeneca R&D Mo¨lndal, S-431 83 Mo¨lndal, Sweden, AstraZeneca R&D Lund, Box 34, S-221 00 Lund, Sweden, and

Research Group for Chemometrics, Umeå University, S-901 87 Umeå, Sweden

Received J anuary 30, 2001

A library of thrombin inhibitors has been designed using statistical molecular design. An

aromatic scaffold was used, with three varied positions corresponding to three pockets at the

active site of thrombin (the S-, P-, and D-pockets). The selection was performed in the building

block space, and previously acquired data were included in the design procedure. The design

resulted in six, four, and six building blocks for the first (S), second (P), and third (D) pockets,

respectively. A second round of selection applied to the combined selected building blocks

resulted in a subset of 18 compounds. The selected library was synthesized in parallel and

biologically evaluated. The compounds were analyzed with respect to their inhibition (pIC₅₀)

of thrombin; membrane permeability, estimated by migration behavior in micellar media (CE

log k′) and pK_a; and specificity with respect to inhibition (K_i) of trypsin. Multivariate QSAR

studies of the responses yielded valuable results and information that could only be found

using statistical molecular design in combination with multivariate analysis.

In tr od u ction

ever, even though combinatorial chemistry expands

synthetic capacity, the number of compounds that can

be synthesized is still limited in practice. Along with

increases in the number of building blocks and positions

varied, the number of possible combinations accelerates

rapidly, and selection is often essential.

The features of combinatorial chemistry, allowing

many compounds to be synthesized in parallel, together

with today’s computer capacity provide an excellent

basis for statistical selection of library compounds. A

smaller subset of molecules can be selected using

statistical experimental techniques, i.e., “statistical

molecular design” (SMD), without losing information

contained in the full library.^18-20The objectives of a

given drug design project, and relevant prior knowledge,

should be the focus of the statistical design and thus

steer the selection process. It is also important to

consider synthetic and financial aspects in the early

stages of the selection procedure.

In library design, there is a choice between basing the

design either on the building blocks or on the enumer-

ated products.^18,19,21,22The former has several advan-

tages;²³e.g., the characterization of physicochemical

properties will be made on a smaller set of compounds

and the number of reactants will be reduced in a

straightforward way. The evaluation process will also

be more direct, indicating not only what molecular

properties should be varied but also in what part of the

molecule appropriate changes should be introduced.

Together with the design of the library of chemical

structures, it is equally important to design a battery

of biological tests for investigating the structures. The

evaluation even at an early stage (preclinical) should

include biologically relevant tests with a broad phar-

macological profile, so the number of compounds to be

discarded in expensive and time-consuming clinical

research phases can be minimized.

Thrombosis is the result of improper regulation of the

hemostasis mechanisms, leading to the formation of

intravascular clots, which may cause tissue damage or

cell death due to inadequate blood flow. This cardio-

vascular disorder can lead to deep vein thrombosis,

myocardial infarction, and stroke. Much current drug

research is focused on finding an antithrombotic drug

that is safe and effective and can be administered orally.

One prominent target for such a drug is the enzyme

thrombin (factor IIa). Thrombin plays a central role in

the blood coagulation process as the final key enzyme

in the cascade involved. It converts soluble fibrinogen

to fibrin, which forms the fibrillar matrix of the blood

clot. In addition, the activation of thrombin initiates

several other mechanisms including both positive and

negative feedback of thrombin generation,¹together

with a variety of cellular effects such as platelet ag-

gregation and tissue remodeling.^2,3A low molecular

weight substance that can inhibit thrombin, and thus

its actions, would be a potentially powerful antithrom-

botic drug. Substances that bind to the active site have

been successfully developed, and several inhibitors have

been described.^4-13However, finding an inhibitor that

comprise both selectivity and suitable pharmacokinetics

has been difficult to identify, prompting the research

to continue.^14-17

In the drug discovery process combinatorial chemistry

is routinely used as a complement to sequential syn-

thesis in medicinal chemistry. This automatic approach

offers the possibility of synthesizing large arrays of

compounds by combining sets of building blocks. How-

* Corresponding author. Tel: + 46-(0)31-706 52 07. Fax: + 46-(0)-

31-776 37 10. E-mail: anna.linusson@astrazeneca.com.

^†AstraZeneca R&D Mo¨lndal.

^‡AstraZeneca R&D Lund.

^§Umeå University.

Published on Web 09/11/2001

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3425

Ch a r a cter iza tion of Str u ctu r es. Descriptors from

both 2D and 3D structures were used. The 2D-based

descriptors were calculated using in-house programmed

calculations.²⁹These descriptors include properties of

size, lipophilicity, polarizability, charge, flexibility, ri-

gidity, and hydrogen-binding capacities, which cover the

relevant latent property structures.^29-31The 3D struc-

tures were generated using Concord,³²and the charac-

terization was made using Spartan software, with the

semiempirical AM1.³³

D-Op t im a l Design . D-optimal designs maximize

det(X′X), where X is the coded model matrix with n rows

(structures) and p terms (model variables) and det

denotes the determinant. Here, these designs were

created to support a model relating the properties of the

compounds and their biological response, y ) X*b + e,

where y is a column with n responses, b is a row vector

containing the coefficients, and e is the residuals. By

maximizing the volume of X′X (the determinant) the

error variance of the coefficients b will be minimized.³⁴

Discrete variables and scaled scores were used as design

variables. The D-optimal designs were augmented by

the molecule closest to the calculated centroid. When

the final library was selected, an additional design

criterion that every building block should be represented

at least twice in the selected set was used along with

the criteria set by the design program.³⁵The D-optimal

designs were generated by MODDE software.³⁵

Da ta An a lytic Meth od s. Two data analytical meth-

ods were used: principal component analysis (PCA) for

compression (summarization) of data sets and partial

least squares projection to latent structures (PLS) to

relate a structure descriptor matrix, X, to an activity

matrix, Y.

The number of significant principal properties of

molecules is often much smaller than the number of

structural descriptors. PCA compresses a data set to its

main “principal” structure, i.e., reexpresses the informa-

tion in the structural descriptors using a smaller

number of new uncorrelated variables. PCA is expressed

in terms of scores (T) and loadings (P ), where X ) 1*xj

+ T*P ′ + E, and xj is a row vector of variable means

and E is the residual matrix.³⁶The number of principal

components was here decided using their eigenvalues

and the interpretation of the loadings.

PLS was used for multivariate quantitative structure-

activity relationship (MQSAR) analysis of both the

historical data and of the library we synthesized and

biologically tested. PLS is a latent variable regression

method which calculates a model taking into account

the variance (in X) together with the correlation be-

tween X and Y; i.e., it finds latent structures in X that

correlate with latent structures in Y.^37,38

The results from PLS can be viewed in score and

loading plots. These plots were used to interpret the

model, along with the model coefficients for the vari-

ables (structural descriptors). The models were vali-

dated by investigation of the explained variation in Y,

R²Y, and the predictive capability, Q²(cross-validation,

in five rounds^39,40). The data set was scaled to zero mean

and unit variance prior to the analyses. All calculations

were made using SIMCA software.⁴¹

F igu r e 1. Schematic representation of the active site of

thrombin.

The theoretical aspects of library design have been

discussed in several papers,^24,25but little real data have

so far been presented and evaluated. In the present

work a statistically designed library of thrombin inhibi-

tors has been synthesized and evaluated through bio-

logical testing and subsequent structure-activity stud-

ies. The design was performed in the building block

space, and since thrombin is a well-known enzyme, both

previously acquired results (“in-house” data) and the

crystal structure were used in the selection procedure.

The selected compounds were synthesized in parallel

in solution, and the bioassessments included their

inhibition of thrombin, their specificity toward thrombin

with respect to trypsin, and their tendency to be

passively absorbed. The resulting designed library and

response values were evaluated using multivariate

methods. In addition, we discuss the use of the present

approach for any size of medicinal compound series.

Ma ter ia l a n d Meth od s

Th e Active Site of Th r om bin . R-Thrombin is a

serine protease and a member of the trypsin family.

Serine proteases are proteolytic enzymes that all contain

the characteristic serine-histidine-aspartate catalytic

triad [Ser195, His57, and Asp102 in thrombin (Bode

numbering²⁶)]. The hydroxyl group of the serine acts as

a nucleophile, targeting the scissile amide bond of the

substrate. There are three principal binding pockets at

the cleaving site, the specificity (S) pocket, the proximal

(P) pocket, and the distal (D) pocket, modeled from the

crystal structure.^26,27The sides of the S-pocket in

thrombin are primarily hydrophobic and at the base of

the site there are a carboxylic acid (Asp¹⁸⁹) and two

backbone carbonyls. In fibrinogen an arginine residue

(P1) coordinates into the S-pocket, forming an ion-ion

interaction with the Asp¹⁸⁹. The P-pocket is hydrophobic,

smaller than the S-pocket, and binds the P2 amino acid

valine of fibrinogen. The D-pocket is also hydrophobic,

larger than the P-pocket, and not so well defined. This

pocket is also called the aryl-binding site and it is

occupied by a phenylalanine in fibrinogen. A schematic

representation of the active site is shown in Figure 1.

In addition to the active site, thrombin has several other

recognition and binding sites,²⁸but these will not be

discussed here.

Biologica l Eva lu a tion . The synthesized compounds

(see Results and Discussion) were subjected to biological

3426 J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21

Linusson et al.

analysis to investigate their capacity to inhibit thrombin

(pIC₅₀), their membrane permeability according to their

migration behavior in micellar media (CE log k′) and

pK_a, and their inhibition (K_i) of trypsin.

Due to the lipophilic character of the library com-

pounds and because these model compounds were

sparingly soluble in buffer, capillary electrophoreses

(CE) was selected as the technique for investigating

membrane permeability. In addition, CE is also more

straightforward than other in vitro techniques, such as

Caco-2 cell permeation, when only minute amounts of

substance are available. Further, the use of micelles as

separation media in capillary electrophoreses (CE) is a

means to experimentally model absorption to a lipid-

like pseudostationary phase. Here, the bile acid sodium

cholate (SC), a native constituent of intestinal fluids,

was used as the micellar electrolyte. The rationale for

this so-called micellar electrokinetic chromatography

(MEKC) method is that separations are mainly due to

differences in molecular size, hydrophobicity, and hy-

drogen bonding, all of which have a strong influence on

in vivo absorption. Notably, correlation has recently

been demonstrated between MEKC data and log P,⁴²

and between MEKC data and in vitro absorption.^43,44

In an MEKC system, micelles are amphiphilic ag-

gregates with an anisotropic microenvironment that

provides both hydrophobic and electrostatic sites of

interactions. In this way, the micellar phase is structur-

ally more similar to biomembranes than 1-octanol/water

partitioning, commonly used for lipophilicity determina-

tions (log P). Moreover, Woodrow and Dorsey⁴⁵have

reported that the thermodynamic signature of micelle-

water partitioning, as used in MEKC, is similar to

biological partitioning.

F igu r e 2. The overall structure of the compounds included

in the MQSAR of the historical data (a) and the common

substructure of the compounds that were designed and syn-

thesized in the present study (b). The substituents for the

previously synthesized compounds (a) were -CH₃, -OCH₃,

-Cl for R′1, with 35 different building blocks for R′2, and the

substituents chosen for synthesis (b) were -H, -CH₃, and

-CH₂CH₃for R1; -H, -OCH₃, and -Cl for R2; -H, -CH₃,

and -Cl for R3; -H and -Cl for R4; -H and -OH for R5; and

-H, -CH₃NH₂, and -CNHNH₂for R6.

groups (without regard to the substituents) consisting

of 8 phenyls, 12 hetereoaromatics, 2 aliphatics, 5 bicyclic

hetereoaromatics, and 8 bicyclic phenyl-phenyls.

The 105 compounds were described by 53 structural

descriptors, including a range of indices, atom counts,

and electronic, lipophilicity, polarizability, and size

parameters, using an “in-house” program.^23,29,31In ad-

dition, binary descriptors were used both to indicate the

three substituents at the P-position (methyl, chloride,

and methoxy) and to describe the five groups repre-

sented at the D-position (phenyl, hetereoaromatic,

aliphatic, bicyclic hetereoaromatic-phenyl, and bicyclic

phenyl-phenyl).

MQSAR modeling of these data, using the structural

descriptors (as the X-block) and the inhibition of throm-

bin (pIC₅₀) as the response, resulted in a significant

model with R²Y ) 0.68, Q²) 0.51 (four components,

using linear terms plus square terms of the continuous

descriptors). The results showed that the methyl group

at the P-residue gave the strongest response. It is

important to note that the building blocks for the

D-residue were not selected according to a statistical

design, making interpretation of these results somewhat

uncertain. However, the outcome indicated that a

benzene ring or a single hetereoaromatic ring might be

preferable at this position.

Bu ild in g Block Selection . Representative sets of

building blocks were selected separately for the three

residues shown in Figure 2, i.e., the S-, P-, and D-

residue. A formal statistical design was used only in the

selection of the D building blocks.

Th e S-Resid u e. In the previously synthesized library

the S-residue was held constant, using a benzamidine

(at position 4) to coordinate to the carboxylic acid in

Asp¹⁸⁹. It is well-known that this group effectively

inhibits thrombin due to the ion-ion interaction it

induces.⁷However, it is also well-known that the

amidine group, due to its basicity, will be protonated

at pH 7, and hence, the rate of its passive absorption in

the body may be low.^15-17Therefore, one objective of the

present study was to investigate if the basicity could

be decreased while the inhibitory effect was maintained.

Because only a minute amount of sample was avail-

able in this study and the thrombin inhibitor compounds

were degraded at neutral to high pH, common tech-

niques for pK_adetermination could not be used. Here,

CE enabled determination of pK_avalues for the 11

compounds containing a basic moiety, although these

were partially degraded. During CE separation, degra-

dation products could be separated from the respective

compound.⁴⁶

Resu lts a n d Discu ssion

Sca ffold . An achiral scaffold was selected for the

library (Figure 2), to which three building blocks were

connected that coordinate to the S-pocket (S-residue),

the P-pocket (P-residue) and the D-pocket (D-residue).

A number of compounds containing this scaffold have

previously been synthesized and tested for thrombin

inhibition,⁴⁷giving the opportunity to include historical

data in the design process. Some of these analogues

have also been cocrystallized with thrombin, confirming

that they bind in the assumed way.⁴⁷

Histor ica l Da ta . One hundred and five previously

synthesized compounds comprising the overall structure

showed in Figure 2a⁴⁷were subjected to MQSAR

analysis. In this library the S-residue was held constant,

while the P-residue (R′₁) was varied with three building

blocks, methyl-, chlorine- and methoxy groups, each

having 35 different D-residues (R′₂, 35 × 3 compounds).

The 35 groups at the D-residue were not varied in a

balanced way. The frameworks could be divided into five

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3427

F igu r e 5. The six building blocks selected for the D-pocket.

synthetic routes showed the phenyl moiety to be ad-

vantageous compared to the hetereoaromatics, and

hence, a benzene ring was selected as the framework

(Figure 2b).

F igu r e 3. The six building blocks selected for the S-pocket.

The positions 2, 3, and 5 were selected for investiga-

tion (R2, R3, and R4), with the following substituents:

R2, -H, -OCH₃, and -Cl; R3, -H, and -Cl; and R4,

-H, -CH₃and -Cl. This choice was partly based on

the numerous commercially available chemicals with

the listed substituents.

The numbers of selected substituents led to 3 × 2 ×

3 ) 18 building blocks for the D-residue. Not all

combinations were necessary for a screening analysis,

so a D-optimal design was applied using discrete design

variables (indicator variables) for the substituents.

Figure 5 shows the six selected compounds.

F igu r e 4. The three building blocks selected for the P-pocket.

A phenyl framework was selected, with two positions

to be varied (R5 and R6). At position 4 (R6) the following

substituents were selected: -H, -CH₂NH₂(primary

amine), and -CHNHNH₂(amidine). When the hydrogen

is used at position 4, the building block will lack a basic

moiety. The primary amine was selected to introduce a

group that was basic, but not as strongly basic as the

amidine moiety. The amidine group was included in the

design as a reference. In addition, the substituents -H

and -OH were investigated at position 3 (R5). A

hydroxy group at position 3 may serve as a hydrogen

donor to the basic groups at position 4 and thereby

decrease pK_afor the protonated forms. The objective was

to see if this could enhance the tendency of the com-

pounds to be passively absorbed. Since interaction

effects between positions 3 and 4 were expected, all

combinations of these five substituents were selected

as building blocks for the S-residue (Figure 3).

Among these six building blocks, S1 and S2 were

commercially accessible, the other four were synthe-

sized.

Th e P -Resid u e. The MQSAR of the historical data

revealed that a polar group like methoxy or an elec-

tronegative group like chlorine was not favorable at this

position (Figure 2a, R′1). Consequently, only aliphatic

residues were selected in this design. An inspection of

the crystal structure indicated that this pocket was

relatively small. Therefore, in this investigation we

selected substituents with a small size range, -H,

-CH₃, and -CH₂CH₃(Figure 4).

Five out of these six were commercially available; D6

was synthesized.

Th e F in a l Libr a r y Selection . A full factorial design

(all combinations) of the selected building blocks (6 × 3

× 6) would result in 108 compounds. However, a

fractional sampling of these should theoretically be

sufficient to investigate their structural space,^19,48and

hence, a smaller subset was selected, based on a

multivariate characterization of the compounds.

The characterization was done in the building block

space.²³The selected building blocks (6 + 3 + 6 ) 15)

were described individually using semiempirical calcu-

lations (AM1), HOMO, LUMO, polarizability, molecular

volume, surface area, log P, dipole moment, maximum

charge (Mulliken), and minimum charge (Mulliken).³³

In these calculations a methyl group was used to replace

the sulfonyl chloride substituent in the reactant set D

and the ethanol substituent in the reactant set S. In

addition, the molecular weight, the number of hydrogen-

bond donors, and acceptors for the S-residue and binary

descriptors for the substituent patterns for residues S

and D were used. Each of the three building block sets

(S, P, and D) was compressed individually by PCA. In

total, five principal properties (PCA scores) were used

as design variables, two for the residues S and D and

one for the P residue. These were combined to represent

the properties of the 108 candidate compounds. The

D-optimal design (linear and quadratic terms) procedure

yielded six different selections of 18 compounds that

were approved according to the design criteria (see

Material and Methods). From those approved selections

the final choice was subjectively made on the basis of

medicinal experience, and Table 1 shows the 18 com-

pounds that were finally synthesized.

Two of these three building blocks were commercially

available, while P3 was synthesized.

Th e D-Resid u e. The MQSAR model of the historical

data indicated that a benzene or hetereoaromatic ring

was favorable at this position. Even though the model

might be somewhat uncertain due to the unbalanced

data, these results were used in the design. Investiga-

tion of commercially available chemicals and possible

Syn th esis of Non com m er cia l S-Bu ild in g Block s.

Both the amidine and amine groups of building blocks

S3-S6 were synthesized from the corresponding nitriles

after synthesizing the library. The nitrile derivative

3428 J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21

Linusson et al.

Ta ble 1. Chemical Structure and Biological Measurement Values for the 18 Selected Compounds

*K_i) IC₅₀/(1 + [S]/K_m); [S] ) 0.3 mM, K_m) 0.33 mM.

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3429

Sch em e 1^a

Table 2). In the first step the P-residue was connected

with the D-residue. Mixing the two building blocks in a

two-phase system consisting of diethyl ether and satu-

rated sodium hydrogen carbonate yielded the mono

sulfonic acid esters 12-26. Increasing the equivalents

of 1,3-dihydroxy benzene derivative compared to the

sulfonyl chloride resulted in higher yields of the mono-

substituted product. In the case of 22-26, this was

necessary to obtain any product at all. In the second

step, the S-residue was connected to the sulfonic acid

esters 12-26 through a Mitsunobu reaction, using

diethylazodicarboxylate (DEAD) and triphenyl phos-

phine in THF, to give 27a -44a . The allyl ethers 27a -

29a , 33a -35a , and 38a -41a were cleaved reductively

with tetrakis(Pd(PPh₃)₄) and sodium borohydride in

THF, yielding 27b -29b , 33b -35b , and 38b -41b .

Reduction of the nitriles 31a , 37a , 43a , 28b, 34b, and

40b using borane-dimethyl sulfide complex in THF

gave the amines 31b, 37b, 43b, 28c, 34c, and 40c.

Finally, the amidines 32b, 44b, 29c, 35c, and 41c were

synthesized through Pinner reaction of 32a , 44a , 29b,

35b, and 41b in ethanol saturated with hydrogen

chloride and subsequent treatment with ammonium-

saturated ethanol. In the synthesis of the library, purity

was favored over yield. The chemical structures and the

purity of the final library were verified and determined

by high-resolution mass spectroscopy (HRMS) and by

proton and carbon nuclear magnetic resonance (¹H

NMR, ¹³C NMR). Two of the final compounds, 36 and

42, did not ionize during the HRMS measurements, and

hence, no masses were recorded for these substances.

a

Reagents and conditions: (a) EtOAc, hexane, NaHSO₄‚SiO₂,

reflux; (b) (1) CH₃CH₂MgBr, Et₂O, 0 °C/rt, (2) paraformaldehyde,

Et₃N, benzene, reflux; (c) CHCl₃, MeOH, K₂CO₂, CH₂CHCH₂Br,

reflux; (d) CH₃CH₂NO₂, NaOAc, AcOH, reflux; (e) K₂CO₃, MeOH,

reflux.

Sch em e 2^a

a

Reagents and conditions: (a) Pd/C, H₂, 4% HCl, rt.

Sch em e 3^a

p IC₅₀of Th r om bin a n d CE log k′. Twelve of the 18

compounds gave a 50% inhibition of thrombin at con-

centrations e133 µM. Eleven of them contained a basic

S-moiety, allowing the possibility of coordination to

Asp¹⁸⁹in the enzyme. The twelfth, which included a S2

S-residue, was only marginally active (pIC₅₀) 4.25).

a

The 11 compounds containing a basic S-moiety were

further used for MQSAR modeling, using two responses,

pIC₅₀for the enzyme activity and CE log k′, as an

estimate of membrane permeability. In this set of 11

compounds the building blocks selected for positions S,

P, and D were represented in a balanced way.

Reagents and conditions: (a) (CH₃)₃SiCHN₂, 60% HBF₄,

CH₂Cl₂, 0 °C.

yielding building blocks S4 and S6, 2-allyloxy-4-(2-

hydroxyethyl)benzonitrile (6), was synthesized in five

steps (Scheme 1). Selective protection of the primary

alcohol of 3-(2-hydroxyethyl)phenol (1) gave 2. Formy-

lation of 2 via the magnesium bromine salt of the phenol

resulted exclusively in the o-hydroxy aldehyde 3. The

phenol of 3 was protected as the allyl ether 4 and the

aldehyde was converted into the nitrile 5 through a one-

pot synthesis using nitroethane and sodium acetate in

acetic acid.⁴⁹Finally, the primary alcohol was depro-

tected, yielding the wanted nitrile derivative 6. In

addition, the phenol of 1 was protected as the allyl ether

7 in a manner similar to that used for 3 (to yield

building block S2). The nitrile derivative yielding build-

ing blocks S3 and S5 was commercially available.

Syn th esis of Non com m er cia l P - a n d D-Bu ild in g

Block s. Reduction of 1-(3,5-dihydroxyphenyl)ethanone

(8) yielded 9 (P3, Scheme 2), and 3,5-dichloro-2-hy-

droxybenzenesulfonyl chloride (10) was converted into

11 through methylation of the hydroxy group using

trimethylsilane diazomethane (D6, Scheme 3).

A modification of the design descriptors was per-

formed to make the MQSAR easier to interpret (Table

3). Using the same characterization as for the design

(semiempirical, AM1³³), the descriptors were divided

into four blocks, according to whether they were based

on size, electronic, hydrophobic, or polarizability pa-

rameters, for each of the varied positions (S, D, and P).

The size and electronic blocks each included more than

one descriptor, and local PCAs were used to reduce the

dimensions (for each building block set, S, D, and P).

The compression resulted in one size descriptor for each

building block set and two electronic properties for the

S- and D-sets, while the P-set only yielded one.

In addition, to allow the effects of the different

substituents at the D-residues to be interpreted, a

semiempirical characterization (AM1) was undertaken

for each of the three varied positions. These structural

descriptors were treated in a manner similar to that

described above, yielding four properties (based on size,

electronic, hydrophobic, and polarizability characteris-

Syn th esis of th e Libr a r y. The library was synthe-

sized in two main steps and five in total (Scheme 4,

3430 J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21

Linusson et al.

Sch em e 4^a

a

Reagents and conditions: (a) Et₂O, aq NaHCO₃, rt; (b) PPh₃, DEAD, THF, rt (c) Pd(PPh₃)₄, NaBH₄, THF, rt; (d) borane-Me₂S, THF,

reflux/rt; (e) (1) EtOH, HCl, rt, (2) EtOH, NH₃, rt. The R-groups are displayed in Table 2.

tics) for each of the three positions (2, 3, and 5) at the

aromatic ring.

(Table 3), with pIC₅₀and CE log k′ as responses (K_X)

87, M_Y) 2), gave four significant components, explain-

ing 98% of the total variation in Y, and a Q²of 0.53 (for

pIC₅₀, R²Y ) 0.97 and Q²) 0.55; for CE log k′, R²Y )

0.98 and Q²) 0.30). The relationship between the

structural properties and the enzyme activity was

nonlinear, as indicated by the large coefficients for the

quadratic term for pK_aand for interaction terms be-

tween the P-residue and the D-residue. Therefore, a

more careful investigation was made of the interaction

effects between the properties of the varied positions

at the D-residue and the properties of the P-residue. A

new MQSAR (K_X) 49, M_Y) 2) was calculated, ex-

panded by the inclusion of two-way interaction terms

between no. 1 and nos. 17-28, while the main terms

(1-31), the cross terms between no. 1 and nos. 12-16

(Table 3) and the quadratic term of pK_a(no 31) were

retained. A refinement of the MQSAR excluding all

nonlinear terms with model coefficients of values in the

range between -0.05 and 0.05 (K_X) 41, M_Y) 2) gave

a model explaining 94% of the total variation and a Q²

of 0.74 (three PLS components: pIC₅₀, R²Y ) 0.91, and

Q²) 0.56; CE log k′, R²Y ) 0.97 and Q²) 0.87). The

terms included in the refined model are listed in Table

Two end product descriptors were added, log D (pH

) 7.4) and log P,⁵⁰where the compounds having the

o-hydroxy amidine were left with missing values, due

to the inaccurate log P estimates. In addition, the

measured pK_avalues were included in the model.

To investigate interaction effects, indicator descriptors

were added to describe building blocks S and P. The

groups in the positions at the S-residue (R₁) were

distinguished by two pairs of variables; at the para

position -1 was set for the primary amine and +1 for

the benzamidine, and at the meta position -1 for the

hydrogen and +1 for the hydroxy group. The P-residue

was varied with -H, -CH₃, and -CH₂CH₃substituents,

represented by -1, 0, and, +1, respectively. The sub-

stitution pattern for the D-residue was more complex

than for the other two and would have generated many

binary descriptors. Therefore, the five principal property

vectors were used to describe this building block. The

structural descriptors used are listed in Table 3 (nos.

1-31).

A PLS using structural descriptors nos. 1-31, square

terms of the continuous descriptors (4-31), and two-

way interaction terms between nos. 1-3 and 12-16

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3431

Ta ble 2. Key to the R-groups in Scheme 4

no.

substituents

no.

substituents

12 R1 ) H, R2 ) Cl, R3 ) Cl, R4 ) H

13 R1 ) H, R2 ) H, R3 ) H, R4 ) H

39a R1 ) Et, R2 ) H, R3 ) H, R4 ) H, R5 ) OCH₂CHCH₂, R6 ) H

40a R1 ) Et, R2 ) Cl, R3 ) Cl, R4 ) H, R5 ) OCH₂CHCH₂, R6 ) CN

41a R1 ) Et, R2 ) OMe, R3 ) Cl, R4 ) Cl, R5 ) OCH₂CHCH₂, R6 ) CN

42 R1 ) Et, R2 ) Cl, R3 ) H, R4 ) Cl, R5 ) H, R6 ) H

43a R1 ) Et, R2 ) OMe, R3 ) H, R4 ) Me, R5 ) H, R6 ) CN

44a R1 ) Et, R2 ) H, R3 ) H, R5 ) H, R6 ) CN

14 R1 ) H, R2 ) OMe, R3 ) H, R4 ) Me

15 R1 ) H, R2 ) H, R3 ) Me, R4 ) H

16 R1 ) H, R2 ) OMe, R3 ) Cl, R4 ) Cl

17 R1 ) Me, R2 ) OMe, R3 ) Cl, R4 ) Cl

18 R1 ) Me, R2 ) Cl, R3 ) H, R4 ) Cl

19 R1 ) Me, R2 ) H, R3 ) Me, R4 ) H

20 R1 ) Me, R2 ) H, R3 ) H, R4 ) H

21 R1 ) Me, R2 ) Cl, R3 ) Cl, R4 ) H

22 R1 ) Et, R2 ) OMe, R3 ) H, R4 ) Me

23 R1 ) Et, R2 ) H, R3 ) H, R4 ) H

24 R1 ) Et, R2 ) Cl, R3 ) Cl, R4 ) H

25 R1 ) Et, R2 ) OMe, R3 ) Cl, R4 ) Cl

26 R1 ) Et, R2 ) Cl, R3 ) H, R4 ) Cl

27a R1 ) H, R2 ) Cl, R3-Cl, R4 ) H, R5 ) OCH₂CHCH₂, R6 ) H

28a R1 ) H, R2 ) H, R3 ) H, R4 ) H, R5 ) OCH₂CHCH₂, R6 ) CN

27b R1 ) H, R2 ) Cl, R3 ) Cl, R4 ) H, R6 ) H

28b R1 ) H, R2 ) H, R3 ) H, F4 ) H, R6 ) CN

29b R1 ) H, R2 ) OMe, R3 ) H, R4 ) Me, R6 ) CN

33b R1 ) Me, R2 ) OMe, R3 ) Cl, R4 ) Cl, R6 ) H

34b R1 ) Me, R2 ) Cl, R3 ) H, R4 ) Cl, R6 ) CN

35b R1 ) Me, R2 ) H, R3 ) Me, R4 ) H, R6 ) CN

38b R1 ) Et, R2 ) OMe, R3 ) H, R4 ) Me, R6 ) H

39b R1 ) Et, R2 ) H, R3 ) H, R4 ) H, R6 ) H

40b R1 ) Et, R2 ) Cl, R3 ) Cl, R4 ) H, R6 ) CN

41b R1 ) Et, R2 ) OMe, R3 ) Cl, R4 ) Cl, R6 ) CN

31b R1 ) H, R2 ) H, R3 ) Me, R4 ) H, R5 ) H

29a R1 ) H, R2 ) OMe, R3 ) H, R4 ) Me, R5 ) OCH₂CHCH₂, R6 ) CN 37b R1 ) Me, R2 ) Cl, R3 ) Cl, R4 ) H, R5 ) H

30 R1 ) H, R2 ) OMe, R3 ) H, R4 ) Me, R5 ) G, R6 ) H

31a R1 ) H, R2 ) H, R3 ) Me, R4 ) H, R5 ) H, R6 ) CN

32a R1 ) H, R2 ) OMe, R3 ) Cl, R4 ) Cl, R5 ) H, R6 ) CN

43b R1 ) Et, R2 ) OMe, R3 ) H, R4 ) Me, R5 ) H

28c R1 ) H, R2 ) H, R3 ) H, R4 ) H, R5 ) H

34c R1 ) Me, R2 ) Cl, R3 ) H, R4 ) Cl, R5 ) OH

33a R1 ) Me, R2 ) OMe, R3 ) Cl, R4 ) Cl, R5 ) OCH₂CHCH₂, R6 ) H 40c R1 ) Et, R2 ) Cl, R3 ) Cl, R4 ) H, R5 ) OH

34a R1 ) Me, R2 ) Cl, R3 ) H, R4 ) Cl, R5 ) OCH₂CHCH₂, R6 ) CN

35a R1 ) Me, R2 ) H, R3 ) Me, R4 ) H, R5 ) OCH₂CHCH₂, R6 ) CN

36 R1 ) Me, R2 ) H, R3 ) H, R4 ) H, R5 ) H, R6 ) H

32b R1 ) H, R2 ) OMe, R3 ) Cl, R4 ) Cl, R5 ) H

44b R1 ) Et, R2 ) H, R3 ) Cl, R4 ) Cl, R5 ) H

29c R1 ) H, R2 ) OMe, R3 ) H, R4 ) Me, R5 ) OH

35c R1 ) Me, R2 ) H, R3 ) Me, R4 ) H, R5 ) OH

37a R1 ) Me, R2 ) Cl, R3 ) Cl, R4 ) H, R5 ) H, R6 ) CN

38a R1 ) Et, R2 ) OMe, R3 ) H, R4 ) Me, R5 ) OCH₂CHCH₂, R6 ) H 41c R1 ) Et, R2 ) OMe, R3 ) Cl, R4 ) Cl, R5 ) OH

Ta ble 3. Terms Used in the Refined MQSAR Model

model

term

no. abbrev

explanation

position term no. abbrev

explanation

linear 22 3•size PCA score, size

linear 23 3•logP lipophilicity

linear 24 3•pol polarizability

position

1

2

3

4

5

6

7

8

9

P

discrete variable: -1, 0, +1

discrete variable: -1, +1

P

S

P

S

D pos 3 linear

BA/A

OH/H

P•elec PCA score, electronic properties

linear 25 5•elec PCA score, electronic properties D pos 5 linear

linear 26 5•size PCA score, size

linear 27 5•logP lipophilicity

linear 28 5•pol

linear 29 logD

linear 30 logP

linear 31 pK_a

P•size

P•pol

PCA score, size

polarizability

D pos 5 linear

global linear

global quadratic

P•logP lipophilicity

polarizability

lipophilicity

basicity

S•elec1 PCA score, electronic proeprties 1

S•elec2 PCA score, electronic properties 2

10 S•size

11 S•pol

PCA score, size

polarizability

S

linear 32 S31×31 pK_a× pK_a

linear 33 C1×12 P × D•elec1

linear 34 C1×14 P × D•size

linear 35 C1×15 P × D•pol

linear 36 C1×16 P × D•logP

linear 37 C1×17 P × 2•elec

12 D•elec1 PCA score, electronic properties 1

13 D•elec2 PCA score, electronic properties 2

14 D•size PCA score, size

D

P × D

interaction

15 D•pol

polarizability

16 D•logP lipophilicity

17 2•elec

18 2•size

PCA score, electronic properties

PCA score, size

D pos 2 linear 38 C1×18 P × 2•size

D pos 2 linear 39 C1×19 P × 2•logP

D pos 2 linear 40 C1×20 P × 2•pol

D pos 2 linear 41 C1×21 P × 3•elec

D pos 3 linear

19 2•logP lipophilicity

20 2•pol

21 3•elec

polarizability

PCA score, electronic properties

a

The abbreviations for nos. 32-41 indicate square (S) and cross (C) terms of the descriptors having the corresponding number.

3, where nos. 1-31 are the main terms and nos. 32-41

are the nonlinear terms.

on the model than the two first components. In the score

plot (Figure 7a) compounds 44b and 32b are located in

the lower right corner, which is the corresponding

location for pIC₅₀in the loading plot (Figure 7b),

indicating that these compounds had good activity but

not very high CE log k′ values. On the other hand,

compounds 40c and 41c are located in the upper right

corner, at the corresponding location of CE log k′,

revealing that these compounds had the opposite prop-

erties. Compound 29c’s location between these pairs

indicated intermediate pIC₅₀and CE log k′ properties.

By using several responses in the same model the design

of compounds that fulfill more than one criterion is

facilitated.

The calculated responses (pIC₅₀and CE log k′), using

the expanded model, versus the experimental values are

shown in Figure 6. A visualization of the score and

loading vectors from the PLS regression shows the

relationship between the compounds, their calculated

structural descriptors, and measured responses. A

comparison of the score and loading plots (Figure 7a,b)

reveals that the first PLS component identified the

compounds that have both low CE log k′ and low

thrombin inhibition (low pIC₅₀), while the second com-

ponent separated the two responses. The third compo-

nent (not shown) had similar appearance to the first

component in the way that pIC₅₀and the CE log k′ were

correlated. However, this component had less influence

A more general interpretation of the model can be

derived using the modeled coefficients (Figure 8). The

3432 J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21

Linusson et al.

F igu r e 6. Calculated response values (using the MQSAR) versus the experimental values for (a) the inhibition of thrombin

expressed as pIC₅₀(R²Y ) 0.91, Q²) 0.56) and (b) the estimated permeability expressed as CE log k′ (R²Y ) 0.97, Q²) 0.87); see

Table 1 for chemical structures.

F igu r e 7. The loading plot (a) and score plot (b) of the MQSAR; for chemical structures see Table 1 and for structural descriptors

see Table 3.

properties of the D-residue by themselves were of minor

importance when describing the enzyme activity (pIC₅₀),

as indicated by their coefficient values, which were close

to zero (Figure 8a). However, the interaction terms

between the P- and D-residues were large, and it was

mainly position 2 at the D-residue that interacted with

the P-residue. The coefficients indicated, for example,

that if a larger building block, like a methyl or ethyl

group, was used at the P-position, a small substituent

(C1×18) with low polarizability (C1×20) and large

electronic property values (C1×17) should be used at

position 2 at the D-residue for increased activity (pIC₅₀).

In the synthesized library the second position at the

D-residue was varied using hydrogen, methoxy, and

chlorine substituents, and thus, hydrogen was shown

to be preferable over the other two in combination with

a methyl or ethyl group at the P-residue. In addition,

the model coefficients for the S-residue showed that

stronger basicity at this position led to stronger inhibi-

tion of thrombin.

In addition, the substituents at position 3 on the

D-residue, had an effect on CE log k′. At this position,

three substituents were varied, hydrogen, methyl, and

chlorine, where chlorine was the substituent that best

matched the properties that, as indicated by the model,

would increase the passive absorption.

p K_a. Consideration of the pK_avalues of the investi-

gated compounds with respect to their inhibition of

thrombin indicated that hydroxylation next to the

amidine (29c, 35c, 41c) introduced nonlinearity, whereby

the pK_awas reduced more than the pIC₅₀(Figure 9,

Table 1).

K_iof Tr yp sin . It was not possible to perform a

MQSAR on the selectivity measurements, since low

solubility prevented the gathering of sufficient experi-

mental data. However, a comparison of the four derived

K_ivalues (trypsin) with the calculated K_ifor thrombin

(from the IC₅₀values) indicated that the introduced

hydroxy group may decrease the trypsin activity and

hence increase specificity toward thrombin (Figure 10,

Table 1).

Evaluation of model coefficients for CE log k′ revealed

that hydroxylation of the S-residue had a positive effect

on the possibilities for molecules to be passively ab-

sorbed, as indicated by several coefficients (Figure 8b).

The present study has shown that using a very small

series of compounds, SMD can enhance the resulting

information content per compound. It is well-known that

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3433

F igu r e 8. The model coefficients from the MQSAR using two responses, (a) the pIC₅₀and (b) the CE log k′ values. A full explanation

of the model terms is given in Table 3.

F igu r e 9. The inhibition of thrombin (pIC₅₀) versus pK_a; for

chemical structures see Table 1.

F igu r e 10. Calculated pK_ifor thrombin versus pK_ifor trypsin;

for chemical structures see Table 1.

scale-up of such series, i.e., the use of parallel or

combinatorial approaches, may cause practical prob-

Con clu sion s

The techniques of statistical experimental design and

parallel synthesis were used to make a library of

thrombin inhibitors. Prior information and synthetic

aspects were included in the design procedure. The

biological evaluation showed that a basic moiety was

needed in this library to inhibit thrombin. Six of the

seven compounds devoid of a basic moiety did not inhibit

thrombin.

The MQSAR studies of thrombin activity (pIC₅₀) and

the tendency of the compounds to be passively absorbed

(CE log k′) yielded valuable results. In the structure-

activity relationship of thrombin inhibition, the model

lems. However, we cannot foresee any other difficulties

and therefore postulate that every compound should

have a medicinal consideration, independent of the

series size the compound is a member of, before it is

made. Thus, we believe that the aforementioned result-

ing information content enhancement is independent of

the size of the compound series and that SMD can be

applied to any size of compound series. However, with

growing series it may be tempting to allow the depth of

medicinal consideration per compound to decrease.

Therefore, the use of moderately sized libraries is

motivated.

3434 J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21

Linusson et al.

chloroform (120 mL) were refluxed for 20 min, followed by

addition of 3 (2 g, 9.6 mmol). Allyl bromide (1.7 g, 14 mmol)

was added after 5 min and the reaction mixture was refluxed

overnight. The solution was poured into ice and extracted three

times with Et₂O. Evaporation under reduced pressure of the

dried organic layer yielded 4 (1.7 g, 85%), which was used

without further purification: ¹H NMR (CDCl₃) δ 2.88 (t, J )

6.3, 2H), 3.87 (t, J ) 6.3, 2H), 4.64 (dt, J ) 5.2, 1.5, 2H), 5.32

(dq, J ) 10.6, 1.5), 5.43 (dq, J ) 17.3, 1.5), 6.00-6.10 (m, 1H),

6.80 (s, 1H), 6.88 (d, J ) 7.9, 1H), 7.76 (d, J ) 7.9, 1H), 10.4

(s, 1H).

Acet ic Acid 2-(3-Allyloxy-4-cya n op h en yl)et h yl E st er

(5). A mixture of freshly distilled nitroethane (1.5 g, 16 mmol),

sodium acetate (2.7 g, 33 mmol), acetic acid (4 mL), and 4 (1.7

g, 8 mmol) was refluxed overnight.⁴⁹The solution was poured

into ice (40 g) and extracted three times with diethyl ether.

The organic layer was washed three times with NaHCO₃

(saturated), dried, and concentrated at reduced pressure.

Purification of the residue by column chromatography (toluene-

ethyl acetate 6:1) gave 5 (1.2 g, 60%): ¹H NMR (CDCl₃) δ 2.01

(s, 3H), 2.94 (t, J ) 6.8, 2H), 4.27 (t, J ) 6.8, 2H), 4.64 (dt, J

) 5.0, 1.5, 2H), 5.31 (dq, J ) 10.6, 1.5, 1H), 5.46 (dq, J ) 17.3,

1.5, 1H), 5.98-6.07 (m, 1H), 6.78 (s, 1H), 6.84 (d, J ) 7.8, 1H),

7.47 (d, J ) 7.8, 1H); IR 1738 (CdO), 2225 (CtN).

2-Allyloxy-4-(2-h yd r oxyeth yl)ben zon itr ile (6). A stirred

solution of potassium carbonate (0.8 g, 5.9 mmol), dry metha-

nol (30 mL), and 5 (1.2 g, 4.9 mmol) was refluxed for 2 h. The

mixture was poured into ice and immediately extracted three

times with diethyl ether. The combined organic layers were

washed with water and brine. Evaporation under reduced

pressure gave 6 (0.77 g, 77%): ¹H NMR (CDCl₃) δ 2.86 (2H, t,

J ) 6.1), 3.78 (2H, t, J ) 6.1), 4.73 (2H, d, J ) 5.1), 5.27-5.31

(1H, m), 5.46-5.52 (1H, m), 6.04-6.14 (1H, m), 6.98 (1H, d, J

) 7.8), 7.11 (1H, s), 7.53 (1H, d, J ) 7.8).

2-(3-Allyloxyp h en yl)eth a n ol (7). The preparation of 7

(quant. yield) was as for 4 except that 1 was used instead of

3: ¹H NMR (CDCl₃) δ 2.72 (2H, t, J ) 6.9), 3.74 (2H, t, J )

6.9), 4.53 (2H, dt, J ) 5.2, 1.7), 5.22 (1H, dq, J ) 10.6, 1.7),

5.40 (1H, dq, J ) 17.3, 1.7), 5.99-6.12 (1H, m), 6.74-6.85 (3H,

m), 7.16 (1H, t, J ) 7.7).

5-Eth ylben zen e-1,3-d iol (9). A reaction mixture of 2 g

(13.1 mmol) of 1-(3,5-Dihydroxyphenyl)ethanone (8), 1 g of

Pd/C (10%), and 150 mL of aqueous HCl (4%) in a 500 mL

round-bottom flask was hydrated overnight (ca 16 h and 900

mL of H₂). The mixture was filtered through Celite and

extracted with three portions of Et₂O. The dried solution was

evaporated at reduced pressure. Purification of the residue by

column chromatography (toluene-ethyl acetate 5:1) gave 9

(1.53 g, 85%) as pale orange crystals: ¹H NMR (CD₃COCD₃)

δ 1.13 (t, J ) 7.6, 3H), 2.45 (q, J ) 7.6, 2H), 6.16 (t, J ) 2.2,

1H), 6.18 (d, J ) 2.2, 2H), 7.98 (s, 1H).

3,5-Dich lor o-2-m eth oxyben zen esu lfon yl Ch lor id e (11).

To a stirred solution of aqueous HBF₄(209 mg, 60%) and CH₂-

Cl₂(40 mL) was added 10 (10 mmol) at 0 °C. Three portions

of TMSCHN₂[2 M hexane solution, 5 mL (10 mmol), 3 mL (6

mmol), 2 mL (4 mmol)] were added dropwise at intervals of

30 min, and the mixture was stirred at 0 °C for an additional

40 min.⁵³The reaction mixture was poured into water and

extracted with two portions of CH₂Cl₂. The organic layer was

washed with water, dried, and evaporated under reduced

pressure. The resulting oil was purified by slow distillation

(85 °C, 0.1 mmHg) and subsequent filtration through SiO₂

(heptane-ethyl acetate 20:1). Removal of the solvent gave 11

(1.65 g, 60%) as an oil that crystallized in the freezer: ¹H NMR

(CDCl₃) δ 4.10 (s, 3H), 7.72 (d, J ) 2.6, 1H), 7.88 (d, J ) 2.6,

1H).

showed strong interaction effects between the properties

of the D-residue and the properties of the P-residue.

This information could only be found using statistical

molecular design in combination with multivariate

analysis.

The hydroxylation of the benzamidine increased the

tendency of the candidate molecules to be passively

absorbed without causing large losses of inhibitory

effectiveness. In addition, this modification also indi-

cated improved selectivity toward thrombin with respect

to trypsin.

To conclude, statistical molecular design is a straight-

forward method that yields maximal and reliable in-

formation from a limited set of molecules, and it can be

combined with synthetic considerations and prior in-

formation.

Exp er im en ta l Section

Gen er a l Ch em istr y. NMR spectra were recorded on a 400

or 600 MHz Varian Unity plus spectrometer. Chemical shifts

(δ) are reported in parts per million (ppm) relative to the

deuterated solvent as indicated in the synthetic procedure

described below. Spin multiplicities are given as s (singlet), d

(doublet), dd (double doublet), ddd (double double doublet), t

(triplet), dt (double triplet), q (quartet), dq (double quartet),

or m (multiplet). Coupling constants (J ) are given in hertz (Hz).

Analytical HPLC was performed using a Waters 717 system

plus Autosampler, with a Symmetry C₈, 5 µm (4.6 · 50 mm)

column, and preparative HPLC was performed using a Prep

LC 2000 (Waters) instrument with a Cromasil C₈, 7 µm, (Akzo

Nobel) column. In both cases the mobile phases were aceto-

nitrile/water gradients (containing 10 mM ammonium acetate).

The UV trace was obtained with a Waters 486 Tunable

Absorbance detector (λ ) 254 nm). High-resolution mass

spectra (HRMS) were produced in a Finnigan MAT900TS

(nanospray) spectrometer using an acetonitrile/water matrix.

All commercially available chemicals were used without

further purification, unless otherwise indicated. Organic ex-

tracts were dried using Na₂SO₄throughout. Column chroma-

tography was performed using Merck silica gel 60 (0.040-

0.063 mm), and infrared spectra were recorded using a

Mattson FTIR spectrometer.

Acetic Acid 2-(3-Hyd r oxyp h en yl)eth yl Ester (2). The

NaHSO₄‚SiO₂catalyst was prepared as described by Breton.⁵¹

A stirred mixture of 1 (2.5 g, 18 mmol), ethyl acetate (80 mL),

hexane (200 mL), and the catalyst (362 mg) was brought to

reflux. After 22 h the solution was cooled to room temperature,

filtered, and concentrated at reduced pressure. Purification of

the residue by column chromatography (toluene-ethyl acetate

5:1) gave 2 (3.1 g, 95%): ¹H NMR (CDCl₃) δ 2.03 (s, 3H), 2.86

(t, J ) 7.0, 2H), 4.26 (t, J ) 7.0, 2H), 6.17 (s, 1H), 6.69-6.71

(m, 3H), 7.14 (t, J ) 7.6, 1H).

Acetic Acid 2-(4-F or m yl-3-h yd r oxyp h en yl)eth yl Ester

(3). Dry THF (40 mL) was added to 2 (2.2 g, 12 mmol). The

mixture was cooled to 0 °C under argon and ethylmagnesium

bromide (3 M THF solution, 12 mmol) was added dropwise.

After removal of the ice bath, the solution was stirred for 30

min and a white salt precipitated. The reaction mixture was

concentrated under reduced pressure, benzene was added in

two portions (10 mL), and the removal of the solvent was

repeated. Benzene (20 mL), paraformaldehyde (2.75 g, 30

mmol), and triethylamine (1.85 g, 18 mmol) were added to the

salt.⁵²The reaction mixture was refluxed for 3 h and then

cooled to room temperature. The solution was made acidic by

HCl (2 M) and was extracted three times with diethyl ether.

The organic layer was dried and evaporated under reduced

pressure to give 3 (2.4 g, 96%), which was used in the next

step without further purification: ¹H NMR (CDCl₃) δ 2.01 (s,

3H), 2.93 (t, J ) 6.8, 2H), 4.28 (t, J ) 6.8, 2H), 6.83-6.87 (m,

2H), 7.47 (d, J ) 7.8, 1H), 9.83 (s, 1H), 11.01 (s, 1H).

Gen er a l P r oced u r e for Com p ou n d s 12-21. The sulfonyl

chloride (3.3 mmol) was added to a stirred solution of the

benzene diol (3 mmol), diethyl ether (4 mL), and NaHCO₃

(saturated, 6 mL). The mixture was kept at room temperature

for 2 days. The organic layer was separated and washed with

two extractions of K₂CO₃(saturated). The product was purified

and isolated by three extractions with NaOH (1 M), and the

water layer was acidified by HCl (2 M), followed by three

2-Allyloxy-4-(2-h yd r oxyeth yl)ben za ld eh yd e (4). Potas-

sium carbonate (5.8 g, 42 mmol), methanol (60 mL), and

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3435

extractions using Et₂O. The combined organic layers were

evaporated under reduced pressure to give the pure monosub-

stitued product.

Ben zen esu lfon ic a cid 3-eth yl-5-h yd r oxyp h en yl ester

(23): yield 38%; ¹H NMR (CD₃COCD₃) δ 1.06 (3H, t, J ) 7.6),

2.46 (2H, q, J ) 7.6), 6.27-6.28 (1H, m), 6.35 (1H, t, J ) 2.2),

7.61-7.62 (1H, m), 7.64-7.68 (2H, m), 7.77-7.81 (1H, m),

7.85-7.88 (2H, m), 8.61 (1H, s, broad).

2,3-Dich lor oben zen esu lfon ic a cid 3-h yd r oxyp h en yl

ester (12): yield 40%; ¹H NMR (CDCl₃) δ 5.51 (1H, s, broad),

6.65-6.72 (3H, m), 7.13 (1H, t, J ) 8.5), 7.30 (1H, t, J ) 8.1),

7.73 (1H, dd, J ) 8.1, 1.5), 7.88 (1H, dd, J ) 8.1, 1.5).

Ben zen esu lfon ic acid 3-h ydr oxyph en yl ester (13): yield

2,3-Dich lor o-ben zen esu lfon ic a cid 3-eth yl-5-h yd r oxy-

1

p h en yl ester (24): yield 29%; H NMR (CD₃COCD₃) δ 1.07

(3H, t, J ) 7.6), 2.49 (2H, q, J ) 7.6), 6.43-6.45 (2H, m), 6.62-

6.63 (1H, m), 7.57 (1H, t, J ) 8.3), 7.94 (1H, dd, J ) 8.3, 1.5),

8.00 (1H, dd, J ) 8.3, 1.5).

1

60%; H NMR (CDCl₃) δ 6.24 (1H, s, broad), 6.46 (1H, ddd, J

) 8.3, 2.4, 0.7), 6.54 (1H, t, J ) 2.4), 6.71 (1H, ddd, J ) 8.3,

2.4, 0.7), 7.06 (1H, t, J ) 8.3), 7.46-7.51 (2H, m), 7.60-7.64

(1H, m), 7.79-7.82 (2H, m).

3,5-Dich lor o-2-m eth oxyben zen esu lfon ic a cid 3-eth yl-

5-h yd r oxyp h en yl ester (25): yield 22%; ¹H NMR (CD₃-

COCD₃) δ 1.09 (3H, t, J ) 7.6), 2.51 (2H, q, J ) 7.6), 4.08 (3H,

s), 7.44-7.46 (2H, m), 7.63-7.64 (1H, m), 7.66 (1H, d, J ) 2.4),

7.96 (1H, d, J ) 2.4), 8.73 (1H, s, broad).

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-h yd r oxy-

1

p h en yl ester (14): yield 40%; H NMR (CDCl₃) δ 2.26 (3H,

s), 3.93 (3H, s), 5.26 (1H, s, broad), 6.62-6.69 (3H, m), 6.94

(1H, d, J ) 8.5), 7.07-7.12 (1H, m), 7.36 (1H, dd, J ) 8.5,

2.0), 7.59 (1H, d, J ) 2.0).

2,5-Dich lor oben zen esu lfon ic a cid 3-eth yl-5-h yd r oxy-

1

p h en yl ester (26): yield 20%; H NMR (CD₃COCD₃) δ 1.08

3-Meth ylben zen esu lfon ic Acid 3-Hyd r oxyp h en yl Ester

(15). The equivalents of benzene diol compared to the sulfonyl

chloride were increased, from 0.9 equiv in the general proce-

(3H, t, J ) 7.6), 2.50 (2H, q, J ) 7.6), 6.44-6.45 (1H, m), 6.46

(1H, t, J ) 2.2), 7.64-7.65 (1H, m), 7.82-7.83 (2H, m), 7.88-

7.89 (1H, m).

1

dure to 3 equiv: yield 73%; H NMR (CDCl₃) δ 2.36 (3H, s),

Gen er a l P r oced u r e for Com p ou n d s 27a -44a . DEAD

(diethylazodicarboxylate, 0.65 mmol) was added dropwise to

a stirred solution of the monosubstituted product (12-26, 0.54

mmol), PPh₃(0.65 mmol), and 2.8 mL of dry THF. The reaction

mixture was kept at room temperature for 16 h, and the

solvent was removed under reduced pressure. The residue was

purified by preparative HPLC to give the products.

2,3-Dich lor oben zen esu lfon ic a cid 3-[2-(3-a llyloxyp h en -

yl)eth oxy]p h en yl ester (27a ): yield 53%; ¹H NMR (CD₃-

COCD₃) δ 3.00 (2H, t, J ) 6.8), 4.15 (2H, t, J ) 6.8), 4.55 (2H,

dt, J ) 5.1, 1.7), 5.23 (1H, dq, J ) 10.5, 1.7), 5.40 (1H, dq, J )

17.3, 1.7), 6.01-6.10 (1H, m), 6.70 (1H, ddd, J ) 8.3, 2.4, 0.7),

6.73 (1H, t, J ) 2.4), 6.81 (1H, ddd, J ) 8.1, 2.7, 1.0), 6.86-

6.91 (3H, m), 7.21 (1H, t, J ) 8.1), 7.25 (1H, t, J ) 8.3), 7.54

(1H, t, J ) 8.1), 7.92-7.97 (2H, m).

Ben zen esu lfon ic a cid 3-[2-(3-a llyloxy-4-cya n op h en yl)-

eth oxy]p h en yl ester (28a ): yield 35%; ¹H NMR (CD₃COCD₃)

δ 3.11 (2H, t, J ) 6.6), 4.19 (2H, t, J ) 6.6), 4.75 (2H, dt, J )

5.2, 1.5), 5.30 (1H, dq, J ) 10.6, 1.5), 5.49 (1H, dq, J ) 17.5,

1.5), 6.03-6.16 (1H, m), 6.56-6.60 (2H, m), 6.85-6.89 (1H,

m), 7.04-7.07 (2H, m), 7.20-7.27 (2H, m), 7.58-7.69 (3H, m),

7.76-7.88 (3H, m).

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-[2-(3-a llyl-

oxy-4-cya n op h en yl)eth oxy]p h en yl ester (29a ): yield 49%;

¹H NMR (CD₃COCD₃) δ 2.24 (3H, s), 3.11 (2H, t, J ) 6.6), 3.97

(3H, s), 4.20 (2H, t, J ) 6.6), 4.20 (2H, t, J ) 6.6), 4.73 (2H,

dt, J ) 5.1, 1.6), 5.28 (1H, dq, J ) 17.3, 1.6), 6.03-6.13 (1H,

m), 6.66-6.69 (2H, m), 6.81-6.84 (1H, m), 7.04 (1H, dd, J )

7.8, 1.5), 7.17-7.23 (3H, m), 7.48-7.53 (2H, m), 7.57 (1H, d, J

) 7.8).

6.18 (1H, s, broad), 6.48 (1H, ddd, (J ) 8.3, 2.4, 1.0), 6.55 (1H,

t, J ) 2.4), 6.71 (1H, ddd, J ) 8.3, 2.4, 1.0), 7.07 (1H, t, J )

8.3), 7.36 (1H, t, J ) 7.8), 7.42 (1H, d, J ) 7.8), 7.60 (1H, d,

7.8), 7.64 (1H, s).

3,5-Dich lor o-2-m eth oxyben zen esu lfon ic acid 3-h ydr oxy-

1

p h en yl ester (16): yield 23%; H NMR (CDCl₃) δ 4.07 (3H,

s), 4.97 (1H, s, broad), 6.66 (1H, t, J ) 2.4), 6.69 (1H, dd, J )

8.3, 2.4), 6.72 (1H, dd, J ) 8.3, 2.4), 7.16 (1H, t, J ) 8.3), 7.64

(1H, d, J ) 2.7), 7.69 (1H, d, J ) 2.7).

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic Acid 3-H y-

d r oxy-5-m eth ylp h en yl Ester (17). The equivalents of ben-

zene diol compared to the sulfonyl chloride were increased,

from 0.9 equiv in the general procedure to 3 equiv: yield 23%;

¹H NMR (CD₃COCD₃) δ 2.80 (3H, s), 4.08 (3H, s), 6.42 (1H, s),

6.47 (1H, s), 6.61 (1H, s), 7.68 (1H, d, J ) 2.4), 7.96 (1H, d, J

) 2.4), 8.64 (1H, s).

2,5-Dich lor oben zen esu lfon ic a cid 3-h yd r oxy-5-m eth -

ylp h en yl ester (18): yield 16%; ¹H NMR (CDCl₃) δ 2.23 (3H,

s), 5.16 (1H, s, broad), 6.43-6.44 (1H, m), 6.53-6.54 (2H, m),

7.52-7.53 (2H, m), 7.93-7.94 (1H, m).

3-Meth ylben zen esu lfon ic Acid 3-Hydr oxy-5-m eth ylph en -

yl Ester (19). The equivalents of benzene diol compared to

the sulfonyl chloride were increased, from 0.9 equiv in the

general procedure to 3 equiv: yield 73%; ¹H NMR (CD₃COCD₃)

δ 2.43 (3H, s), 3.04 (3H, s), 6.31-6.33 (2H, m), 6.59-6.60 (2H,

m), 7.52 (1H, t, J ) 7.6), 7.59 (1H, d, J ) 7.6), 7.65 (1H, d, J

) 7.6), 7.69 (1H, s), 8.60 (1H, s).

Ben zen esu lfon ic a cid 3-h yd r oxy-5-m eth ylp h en yl ester

1

(20): yield 73%; H NMR (CDCl₃) δ 2.19 (3H, s), 5.16 (1H, s),

6.28-6.30 (1H, m), 6.36-6.37 (1H, m), 6.51-6.52 (1H, m),

7.49-7.53 (2H, m), 7.62-7.67 (1H, m), 7.83-7.85 (2H, m).

2,3-Dich lor oben zen esu lfon ic a cid 3-h yd r oxy-5-m eth -

ylp h en yl ester (21): yield 20%; ¹H NMR (CDCl₃) δ 2.22 (1H,

s), 5.20 (1H, s, broad), 6.41-6.43 (1H, m), 6.52-6.53 (2H, m),

7.31 (1H, t, J ) 8.1), 7.73 (1H, dd, J ) 8.1, 1.7), 7.9 (1H, dd,

J ) 8.1, 1.7).

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-p h en eth yl-

oxyp h en yl ester (30): yield 84%; ¹H NMR (CD₃COCD₃) δ 2.25

(3H, s), 3.03 (2H, t, J ) 6.8), 3.97 (3H, s), 4.13 (2H, t, J ) 6.8),

6.65-6.69 (2H, m), 6.83 (1H, dd, J ) 5.8, 2.4), 7.18-7.23 (3H,

m), 7.29-7.31 (4H, m), 7.49-7.53 (2H, m); HRMS calcd for

C

₂₂H₂₂O₅S, 398.1188; observed, 398.1188.

Tolu en e-3-su lfon ic a cid 3-[2-(4-cya n op h en yl)eth oxy]-

Gen er a l P r oced u r e for Com p ou n d s 22-26. The sulfonyl

chloride (3 mmol) was added to a stirred solution of the

benzene diol (9 mmol), diethyl ether (15 mL), and NaHCO₃

(saturated, 15 mL). The mixture was kept at room temperature

for 2 days. The organic layer was separated and concentrated

under reduced pressure. Ts-trisamine (1 g, 3.66 mmol) and

CH₂Cl₂(7 mL) were added, and the solution was kept at room

temperature for 2.5 h. The mixture was filtered, and the resins

were sequentially washed with CH₂Cl₂and Et₂O three times.

The solvent was removed under reduced pressure, and the

residue was purified by preparative HPLC to give the mono-

substituted product.

2-Met h oxy-5-m et h ylben zen esu lfon ic a cid 3-et h yl-5-

h yd r oxyp h en yl ester (22): yield 35%; ¹H NMR (CD₃COCD₃)

δ 1.08 (3H, t, J ) 7.6), 2.27 (3H, s), 2.48 (2H, q, J ) 7.6), 3.99

(3H, s), 6.41-6.42 (1H, m), 6.45 (1H, t, J ) 2.2), 6.57-6.58

(1H, m), 7.21 (1H, d, J ) 8.1), 7.54-7.50 (2H, m).

1

p h en yl ester (31a ): yield 49%; H NMR (CD₃COCD₃) δ 2.41

(3H, s), 3.14 (2H, t, J ) 6.6), 4.19 (2H, t, 6.6), 6.57-6.61 (2H,

m), 7.23 (1H, t, J ) 8.3), 7.49-7.72 (8H, m).

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic a cid 3-[2-(4-

cya n op h en yl)eth oxy]p h en yl ester (32a ): yield 25%; ¹H

NMR (CD₃COCD₃) δ 3.17 (2H, t, J ) 6.6), 4.07 (3H, s), 4.23

(2H, t, J ) 6.6), 6.69-6.73 (2H, m), 6.89 (1H, ddd, J ) 8.5,

2.2, 1.0), 7.25-7.29 (1H, m), 7.53-7.56 (2H, m), 7.66 (1H, d, J

) 2.7), 7.70-7.73 (2H, m), 7.95 (1H, d, J ) 2.7).

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic a cid 3-[2-(3-

a llyloxyp h en yl)eth oxy]-5-m eth ylp h en yl ester (33a ): yield

25%; ¹H NMR (CD₃COCD₃) δ 2.24 (3H, s), 3.30 (2H, t, J )

6.8), 4.07 (3H, s), 4.12 (2H, t, J ) 6.8), 4.56 (2H, dt, J ) 5.1,

1.7), 5.23 (1H, dq, J ) 10.7, 1.7), 5.40 (1H, dq, J ) 17.3, 1.7),

6.02-6.11 (1H, m), 6.5 (1H, t, J ) 2.2), 6.56-6.57 (1H, m),

6.71-6.72 (1H, m), 6.81 (1H, d, J ) 7.8), 6.87 (1H, d, J ) 7.8),

3436 J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21

Linusson et al.

6.90-6.91 (1H, m), 7.21 (1H, t, J ) 7.8), 7.67 (1H, d, J ) 2.7),

7.92 (1H, d, J ) 2.7).

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-[2-(4-cy-

a n op h en yl)eth oxy]-5-eth ylp h en yl ester (43a ): yield 32%;

¹H NMR (CD₃COCD₃) δ 1.07 (3H, t, J ) 7.6), 2.25 (3H, s), 2.51

(2H, q, J ) 7.6), 3.13 (2H, t, J ) 6.6), 3.97 (3H, s), 4.18 (2H, t,

J ) 6.6), 6.48-6.51 (2H, m), 6.66-6.68 (1H, m), 7.20 (1H, d, J

) 8.4), 7.68-7.71 (2H, m).

Ben zen esu lfon ic a cid 3-[2-(4-cya n op h en yl)eth oxy]-5-

eth ylp h en yl ester (44a ): yield 44%; ¹H NMR (CD₃COCD₃) δ

1.05 (3H, t, J ) 7.6), 2.49 (2H, q, J ) 7.6), 3.12 (2H, t, J )

6.6), 4.16 (2H, t, J ) 6.6), 6.39-6.40 (1H, m), 6.42 (1H, t, J )

2.2), 6.70-6.71 (1H, m), 7.50-7.52 (2H, m), 7.62-7.70 (4H,

m), 7.75-7.80 (1H, m), 7.85-7.88 (2H, m).

Gen er a l P r oced u r e for Com p ou n d s 27b-29b, 33b-

35b, a n d 38b-41b. Tetrakis(Pd(PPh₃)₄(8.3 µmol) was added

to a solution of 27a (42 µmol) and 0.44 mL of dry THF. After

5 min, NaBH₄(63 µmol) was added to the stirred solution and

the reaction mixture was kept at room temperature for 3 h.

The reaction was quenched with 1 M HCl (to neutral pH). THF

was removed under reduced pressure, and the aqueous phase

was extracted three times with CH₂Cl₂. The solvent was

removed under reduced pressure and the residue was purified

by preparative HPLC to give the deprotected products.

2,3-Dich lor oben zen esu lfon ic acid 3-[2-(3-h ydr oxyph en -

yl)eth oxy]p h en yl ester (27b): yield 80%; ¹H NMR (CD₃OD)

δ 2.95 (2H, t, J ) 6.8), 4.07 (2H, t, J ) 6.8), 6.64 (1H, t, J )

2.2), 6.67-6.75 (4H, m), 6.84 (1H, dd, J ) 8.3, 1.0), 7.11-7.16

(1H, m), 7.22 (1H, t, J ) 8.3), 7.44 (1H, t, J ) 8.1), 7.87-7.91

(2H, m); HRMS calcd for C₂₀H₁₆Cl₂O₅S, 438.0095; observed,

438.0095.

2,5-Dich lor oben zen esu lfon ic a cid 3-[2-(3-a llyloxy-4-

cya n op h en yl)eth oxy]-5-m eth ylp h en yl ester (34a ): yield

48%; ¹H NMR (CD₃COCD₃) δ 2.22 (3H, s), 3.10 (2H, t, J )

6.6), 4.19 (2H, t, J ) 6.6), 4.73 (2H, dt, J ) 5.1, 1.7), 5.29 (1H,

dq, J ) 10.5, 1.7), 5.49 (1H, dq, J ) 17.3, 1.7), 6.04-6.13 (1H,

m), 6.56-6.57 (1H, m), 6.71-6.72 (1H, m), 7.03 (1H, dd, J )

7.8, 1.5), 7.18 (1H, d, J ) 1.2), 7.57 (1H, d, J ) 7.8), 7.79 (2H,

d, J ) 1.5), 7.89 (1H, t, J ) 1.5).

Tolu en e-3-su lfon ic a cid 3-[2-(3-a llyloxy-4-cya n op h en -

yl)eth oxy]-5-m eth ylp h en yl ester (35a ): yield 35%; ¹H NMR

(CD₃COCD₃) δ 2.21 (3H, s), 2.41 (3H, s), 3.08 (2H, t, J ) 6.6),

4.15 (2H, t, J ) 6.6), 4.74 (2H, dt, J ) 5.1, 1.6), 5.29 (1H, dq,

J ) 10.5, 1.6), 5.49 (1H, dq, J ) 17.3, 1.6), 6.04-6.14 (1H, m),

6.38 (1H, t, J ) 2.2), 6.43-6.44 (1H, m), 6.68-6.69 (1H, m),

7.03 (1H, dd, J ) 7.8, 1.2), 7.18 (1H, d, J ) 1.2), 7.51 (1H, t, J

) 7.8), 7.57 (2h, d, J ) 7.8), 7.63-7.66 (1H, m), 7.68-7.69 (1H,

m).

Ben zen esu lfon ic a cid 3-m eth yl-5-p h en eth yloxyp h en yl

ester (36): yield 60%; ¹H NMR (CD₃COCD₃) δ 2.20 (3H, s),

2.99 (2H, t, J ) 6.8), 4.06 (2H, t, J ) 6.8), 6.37 (1H, t, J )

2.2), 6.45-6.46 (1H, m), 6.68-6.69 (1H, m), 7.19-7.32 (5H,

m), 7.60-7.65 (2H, m), 7.72-7.77 (1H, m), 7.86-7.89 (2H, m);

HRMS calcd for C₂₁H₂₀O₄S, 368.1082; observed, no ions.

2,3-Dich lor oben zen esu lfon ic a cid 3-[2-(4-cya n op h en -

yl)eth oxy]-5-m eth ylp h en yl ester (37a ): yield 42%; ¹H NMR

(CD₃COCD₃) δ 2.21 (3H, s), 3.13 (2H, t, J ) 6.6), 4.18 (2H, t,

J ) 6.6), 6.49-6.51 (1H, m), 6.56-6.57 (2H, m), 6.69-6.70 (1H,

m), 7.49-7.52 (2H, m), 7.56 (1H, t, J ) 8.1), 7.67-7.70 (2H,

m), 7.93-7.98 (2H, m).

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-[2-(3-a llyl-

oxyp h en yl)eth oxy]-5-eth ylp h en yl ester (38a ): yield 27%;

¹H NMR (CD₃COCD₃) δ 1.08 (3H, t, J ) 7.6), 2.24 (3H, s), 2.52

(2H, q, J ) 7.6), 2.99 (2H, t, J ) 6.8), 3.97 (3H, s), 4.11 (2H, t,

J ) 6.8), 4.55 (2H, dt, J ) 5.1, 1.7), 5.23 (1H, dq, J ) 10.7,

1.7), 5.41 (1H, dq, J ) 17.3, 1.7), 6.02-6.11 (1H, m), 6.49-

6.52 (2H, m), 6.67-6.68 (1H, m), 6.81 (1H, ddd, J ) 8.3, 2.4,

1.0), 6.85-6.88 (1H, m), 6.90-6.91 (1H, m), 7.17-7.23 (2H,

m), 7.47-7.50 (1H, m), 7.53-7.54 (1H, m).

Ben zen esu lfon ic a cid 3-[2-(3-a llyloxyp h en yl)eth oxy]-

5-eth ylp h en yl ester (39a ): yield 44%; ¹H NMR (CD₃COCD₃)

δ 2.98 (2H, t, J ) 6.8), 4.10 (2H, t, J ) 6.8), 4.56 (2H, dt, (J )

5.1, 1.7), 5.23 (1H, dq, J ) 10.5, 1.7), 5.41 (1H, dq, J ) 17.3,

1.7), 6.02-6.12 (1H, m), 6.39-6.50 (1H, m), 6.71-6.72 (1H,

m), 6.81 (1H, ddd, J ) 8.3, 2.7, 1.0), 6.85-6.88 (1H, m), 6.90-

6.91 (1H, m), 7.21 (1H, t, J ) 7.8), 7.61-7.65 (2H, m), 7.74-

7.78 (1H, m), 7.85-7.88 (2H, m).

Ben zen esu lfon ic a cid 3-[2-(4-cya n o-3-h yd r oxyp h en yl)-

1

eth oxy]p h en yl ester (28b): yield 60%; H NMR (CD₃OD) δ

2.99 (2H, t, J ) 6.6), 4.09 (2H, t, J ) 6.6), 6.51 (1H, t, J )

2.2), 6.58 (1H, dd, J ) 8.1, 1.7), 6.81-6.89 (3H, m), 7.20 (1H,

t, J ) 8.1), 7.43 (1H, d, J ) 7.8), 7.60 (2H, t, J ) 7.8), 7.74

(1H, t, J ) 7.6), 7.84 (2H, d, J ) 7.8).

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-[2-(4-cy-

a n o-3-h yd r oxyp h en yl)eth oxy]p h en yl ester (29b): yield

1

71%; H NMR (CD₃COCD₃) δ 2.26 (3H, s), 3.04 (2H, t, J )

6.6), 3.98 (3H, s), 4.16 (2H, t, J ) 6.6), 6.65-6.69 (2H, m), 6.82

(1H, ddd, J ) 8.3, 2.4, 1.0), 6.95 (1H, dd, J ) 7.8, 1.5), 7.04

(1H, d, J ) 1.0), 7.19-7.24 (2H, m), 7.49-7.53 (3H, m).

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic a cid 3-[2-(3-

h yd r oxyp h en yl)eth oxy]-5-m eth ylp h en yl ester (33b): yield

1

32%; H NMR (CD₃SOCD₃) δ 2.21 (3H, s), 2.87 (2H, t, J )

6.8), 3.99 (3H, s), 4.05 (2H, t, J ) 6.8), 6.42 (1H, t, J ) 2.2),

6.52-6.53 (1H, m), 6.60-6.62 (1H, m), 6.65-6.67 (2H, m),

6.74-6.75 (1H, m), 7.07 (1H, t, J ) 7.9), 7.70 (1H, d, J ) 2.6),

8.19 (1H, d, J ) 2.6), 9.31 (1H, s, broad); HRMS calcd for

C

₂₂H₂₀Cl₂O₆S, 482.0358; observed, 482.0330.

2,3-Dich lor oben zen esu lfon ic a cid 3-[2-(3-a llyloxy-4-

cyan oph en yl)eth oxy]-5-eth ylph en yl ester (40a): yield 32%;

¹H NMR (CD₃COCD₃) δ 1.06 (3H, t, J ) 7.6), 2.52 (2H, q, J )

7.6), 3.11 (2H, t, J ) 6.6), 4.20 (2H, t, J ) 6.6), 4.74 (2H, dt, J

) 5.1, 1.7), 5.29 (1H, dq, J ) 10.5, 1.7), 5.48 (1H, dq, J ) 17.3,

1.7), 6.04-6.13 (1H, m), 6.52-6.53 (1H, m), 6.55 (1H, t, J )

2.2), 6.72-6.73 (1H, m), 7.04 (1H, dd, J ) 7.8, 1.5), 7.18 (1H,

d, J ) 1.2), 7.54-7.58 (2H, m), 7.94 (1H, dd, J ) 8.1, 1.5),

7.97 (1H, dd, J ) 8.1, 1.5).

2,5-Dich lor oben zen esu lfon ic a cid 3-[2-(4-cya n o-3-h y-

d r oxyp h en yl)eth oxy]-5-m eth ylp h en yl ester (34b): yield

1

54%; H NMR (CD₃OD) δ 2.27 (3H, s), 3.03 (2H, t, J ) 6.3),

4.13 (2H, t, J ) 6.3), 6.45-6.47 (1H, m), 6.55-6.56 (1H, m),

6.71-6.72 (1H, m), 6.87 (1H, dd, J ) 8.1, 1.2), 6.91 (1H, d, J

) 1.2), 7.46 (1H, d, J ) 8.1), 7.74-7.75 (2H, m), 7.91 (1H, dd,

J ) 1.7, 1.0).

Tolu en e-3-su lfon ic a cid 3-[2-(4-cya n o-3-h yd r oxyp h en -

yl)eth oxy]-5-m eth ylp h en yl ester (35b): yield 78%; ¹H NMR

(CDCl₃) δ 2.17 (3H, s), 2.37 (3H, s), 2.93 (2H, t, J ) 6.6), 4.00

(2H, t, J ) 6.6), 5.87 (1H, s, broad), 6.32-6.35 (1H, m), 6.54

(1H, s), 6.77 (1H, d, J ) 8.1), 6.87 (1H, s), 7.34-7.44 (3H, m),

7.60 (1H, d, J ) 7.6), 7.64 (1H, s).

2-Meth oxy-5-m eth ylben zen esolfon ic a cid 3-eth yl-5-[2-

(3-h yd r oxyp h en yl)eth oxy]p h en yl ester (38b): yield 55%;

¹H NMR (CD₃OD) δ 1.14 (3H, t, J ) 7.8), 2.27 (3H, s), 2.55

(2H, q, J ) 7.6), 2.95 (2H, t, J ) 6.8), 3.98 (3H, s), 4.05 (2H, t,

J ) 6.8), 6.41 (1H, t, J ) 2.2), 6.50-6.51 (1H, m), 6.66-6.76

(4H, m), 7.12-7.18 (2H, m), 7.48-7.54 (2H, m); HRMS calcd

for C₂₄H₂₆O₆S, 442.1450; observed, 442.1442.

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic a cid 3-[2-(3-

allyloxy-4-cyan oph en yl)eth oxy]-5-eth ylph en yl ester (41a):

1

yield 17%; H NMR (CD₃COCD₃) δ 1.09 (3H, t, J ) 7.6), 2.54

(2H, q, J ) 7.6), 3.12 (2H, t, J ) 6.6), 4.07 (3H, s), 4.21 (2H, t,

J ) 6.6), 4.75 (2H, dt, J ) 5.1, 1.7), 5.29 (1H, dq, J ) 10.7,

1.7), 5.49 (1H, dq, J ) 17.3, 1.7), 6.03-6.14 (1H, m), 6.53-

6.55 (2H, m), 6.73-6.74 (1H, m), 7.06 (1H, dd, J ) 7.8, 1.2),

7.20 (1H, d, J ) 1.2), 7.59 (1H, d, J ) 7.8), 7.65 (1H, d, J )

2.5), 7.95 (1H, d, J ) 2.5).

2,5-Dich lor oben zen esu lfon ic a cid 3-eth yl-5-p h en eth yl-

oxyp h en yl ester (42): yield 18%; ¹H NMR (CD₃COCD₃) δ 1.09

(3H, t, J ) 7.6), 2.54 (2H, q, J ) 7.6), 3.04 (2H, t, J ) 6.8),

4.15 (2H, t, J ) 6.8), 6.53-6.86 (2H, m), 6.75-6.77 (1H, m),

7.20-7.25 (1H, m), 7.30-7.32 (4H, m), 7.80-7.81 (2H, m),

7.89-7.90 (1H, m); HRMS calcd for C₂₂H₂₀Cl₂O₄S, 450.0459;

observed, no ions.

Ben zen esu lfon ic a cid 3-eth yl-5-[2-(3-h yd r oxyp h en yl)-

eth oxy]p h en yl ester (39b): yield 34%; ¹H NMR (CD₃COOD)

δ 1.07 (3H, t, J ) 7.6), 2.49 (2H, q, J ) 7.6), 2.93 (2H, t, J )

6.8), 4.02 (2H, t, J ) 6.8), 6.35 (1H, t, J ) 2.2), 6.37-6.38 (1H,

m), 6.64-6.65 (1H, m), 6.70-6.77 (3H, m), 7.12 (1H, t, J )

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3437

7.8), 7.54 (2H, t, J ) 7.3), 7.67 (1H, t, J ) 7.3), 7.84 (2H, d, J

J ) 7.7), 2.97 (2H, t, J ) 6.7), 4.08 (2H, t, J ) 6.7), 4.21 (2H,

s), 6.52-6.53 (2H, m), 6.66 (1H, s), 6.80 (1H, d, J ) 7.7), 6.88

(1H, s), 7.20 (1H, d, J ) 7.7), 7.40 (1H, t, J ) 8.1), 7.83 (1H, d,

J ) 8.1), 7.90 (1H, d, J ) 8.1); HRMS calcd for C₂₃H₂₃Cl₂NO₅S,

495.0674; observed, 495.0667.

Gen er a l P r oced u r e for Com p ou n d s 32b a n d 44b. A

solution of ethanol saturated with HCl (2.5 mL) and 32a (46

µmol) was kept at room temperature for 2 days. The solvent

was removed under reduced pressure and 3 mL of ethanol

saturated with NH₃was added. The reaction mixture was

stirred at room temperature for 2 days and the solvent was

again removed under reduced pressure. The residue was

purified by preparative HPLC to yield the product.

) 7.3); HRMS calcd for

C₂₂H₂₂O₅S, 398.1188; observed,

398.1193.

2,3-Dich lor oben zen esu lfon ic a cid 3-[2-(4-cya n o-3-h y-

dr oxyph en yl)eth oxy]-5-eth ylph en yl ester (40b): yield 67%;

¹H NMR (CD₃OD) δ 1.09 (3H, t, J ) 7.8), 2.52 (2H, q, J ) 7.8),

3.03 (2H, t, J ) 6.3), 4.11 (2H, t, J ) 6.3), 6.47 (1H, t, J )

2.2), 6.49-6.50 (1H, m), 6.67-6.68 (1H, m), 6.85 (1H, d, J )

8.1), 6.91 (1H, s), 7.42-7.47 (2H, m), 7.87-7.90 (2H, m).

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic a cid 3-[2-(4-

cyan o-3-h ydr oxyph en yl)eth oxy]-5-eth ylph en yl ester (41b):

1

yield 52%; H NMR (CD₃COCD₃) δ 1.09 (3H, t, J ) 7.6), 2.54

(2H, q, J ) 7.6), 3.04 (2H, t, J ) 6.6), 4.07 (3H, s), 4.16 (2H, t,

J ) 6.6), 6.53 (2H, d, J ) 1.7), 6.73 (1H, t, J ) 1.7), 6.94 (1H,

dd, J ) 7.8, 1.2), 7.05 (1H, d, J ) 1.2), 7.50 (1H, d, J ) 7.8),

7.65 (1H, d, J ) 7.8), 7.95 (1H, d, J ) 2.7).

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic a cid 3-[2-(4-

ca r ba m im id oylp h en yl)eth oxy]p h en yl ester (32b): yield

1

36%; H NMR (CD₃COOD) δ 3.18 (2H, t, J ) 6.4), 4.07 (3H,

Gen er a l P r oced u r e for Com p ou n d s 31b, 37b, a n d 43b.

A solution of 31a (65 µmol) and dry THF (1 mL) was brought

to reflux and BMS (borane-Me₂S, 129 µmol) was added

dropwise. Two additional portions of BMS (2 × 129 µmol) were

added after 30 min and 1 h, respectively, and the reaction

mixture was refluxed for an additional 2.5 h. The condenser

was removed and methanol was added and subsequently

removed by heating the mixture. This procedure was repeated

twice before total removal of the solvent under reduced

pressure. The residue was purified by preparative HPLC to

yield the product.

Tolu en e-3-su lfon ic a cid 3-[2-(4-a m in om eth ylp h en yl)-

eth oxy]p h en yl ester (31b): yield 55%; ¹H NMR (CD₃COCD₃)

δ 2.4 (3H, s), 2.98 (2H, t, J ) 6.8), 4.08 (2H, t, J ) 6.8), 4.42

(2H, s), 6.59 (1H, ddd, J ) 8.3, 2.4, 1.0), 6.86 (1H, ddd, J )

8.3, 2.4, 1.0), 7.21-7.29 (5H, m), 7.50 (1H, t, J ) 7.8), 7.58

(1H, d, J ) 7.8), 7.63 (1H, d, J ) 7.8), 7.67 (1H, s); HRMS

calcd for C₂₂H₂₃NO₄S, 397.1348; observed, 397.1345.

2,3-Dich lor oben zen esu lfon ic a cid 3-[2-(4-a m in om eth -

ylp h en yl)eth oxy]-5-m eth ylp h en yl ester (37b): yield 50%;

¹H NMR (CD₃COOD) δ 2.21 (3H, s), 3.03 (2H, t, J ) 6.4), 4.09

(2H, t, J ) 6.4), 4.23 (2H, s), 6.48 (1H, s), 6.55 (1H, s), 6.64

(1H, s), 7.29-7.45 (5H, m), 7.84 (1H, d, J ) 8.1), 7.91 (1H, d,

J ) 8.1); HRMS calcd for C₂₂H₂₁Cl₂NO₄S, 465.0568; observed,

465.0567.

s), 6.21 (2H, t, J ) 6.4), 6.69 (1H, dd, J ) 2.2, 1.0), 6.72-6.73

(2H, m), 6.82-6.86 (1H, m), 7.22 (1H, t, J ) 8.1), 7.53 (2H, d,

J ) 8.6), 7.67 (1H, d, J ) 2.6), 7.78 (1H, d, J ) 2.6), 7.83 (2H,

d, J ) 8.6); HRMS calcd for

C₂₂H₂₀Cl₂N₂O₅S, 494.0470;

observed, 494.0471.

Ben zen esu lfon ic a cid 3-[2-(4-ca r ba m im id oylp h en yl)-

eth oxy]-5-eth ylp h en yl ester (44b): yield 37%; ¹H NMR

(CD₃COOD) δ 1.06 (3H, t, J ) 7.6), 2.42 (2H, q, J ) 7.6), 3.15

(2H, t, J ) 6.3), 4.16 (2H, t, J ) 6.3), 6.33 (1H, s), 6.45 (1H, t,

J ) 2.2), 6.65 (1H, s), 7.50-7.59 (4H, m), 7.70 (1H, t, J ) 7.3),

7.84 (4H, t, J ) 8.5); HRMS calcd for C₂₃H₂₄N₂O₄S, 424.1457;

observed, 424.1458.

Gen er a l P r oced u r e for Com p ou n d s 29c, 35c, a n d 41c.

A solution of ethanol saturated with HCl (3 mL) and 29b (121

µmol) was kept at room temperature for 5 days. The solvent

was removed under reduced pressure and 3 mL of ethanol

saturated with NH₃was added. The reaction mixture was

stirred at room temperature for 6 h and the solvent was again

removed under reduced pressure. The residue was purified by

preparative HPLC to yield the product.

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-[2-(4-ca r -

ba m im id oyl-3-h yd r oxyp h en yl)eth oxy]p h en yl ester (29c):

1

yield 34%; H NMR (CD₃COOD) δ 2.23 (3H, s), 3.04 (2H, t, J

) 6.1), 3.94 (3H, s), 4.15 (2H, t, J 06.1), 6.65 (1H, dd, J ) 8.3,

2.2), 6.71 (1H, t, J ) 2.2), 6.78 (1H, dd, J ) 8.3, 2.2), 6.97 (1H,

d, J ) 8.3), 7.06-7.18 (3H, m), 7.42 (1H, d, J ) 8.3), 7.56 (1H,

s), 7.70 (1H, d, J ) 8.3); HRMS calcd for C₂₃H₂₄N₂O₆S,

456.1355; observed, 456.1362.

2-Meth oxy-5-m eth ylben zen esu lfon ic a cid 3-[2-(4-a m i-

n om eth ylp h en yl)eth oxy]-5-eth ylp h en yl ester (43b): yield

1

25%; H NMR (CD₃COOD) δ 1.08 (3H, t, J ) 7.6), 2.24 (3H,

s), 2.50 (2H, q, J ) 7.6), 3.04 (2H, t, J ) 6.6), 3.95 (3H, s), 4.11

(2H, t, J ) 6.6), 4.20 (2H, s), 6.48-6.49 (1H, m), 6.51 (1H, t, J

) 2.2), 6.61-6.62 (1H, m), 7.07 (1H, d, J ) 8.5), 7.32 (2H, d,

J ) 8.3), 7.39-7.44 (3H, m), 7.57 (1H, d, J ) 2.2); HRMS calcd

for C₂₅H₂₉NO₅S, 455.1766; observed, 455.1771.

Tolu en e-3-su lfon ic a cid 3-[2-(4-ca r ba m im id oyl-3-h y-

d r oxyp h en yl)eth oxy]-5-m eth ylp h en yl ester (35c): yield

29%; ¹H NMR (CD₃COOD) δ 2.19 (3H, s), 2.38 (3H, s), 3.02

(2H, t, J ) 6.4), 4.09 (2H, t, J ) 6.4), 6.37-6.39 (2H, m), 6.63

(1H, s), 6.86 (1H, dd, J ) 8.4, 1.2), 7.09 (1H, d, J ) 1.2), 7.40-

7.52 (2H, m), 7.62-7.73 (3H, m); HRMS calcd for C₂₃H₂₄N₂O₅S,

440.1406; observed, 440.1406.

Gen er a l P r oced u r e for Com p ou n d s 28c, 34c, a n d 40c.

BMS (81 µmol) was added dropwise to a stirred solution of

28b (63 µmol) and dry THF (1 mL). Two additional portions

of BMS (2 × 81 µmol) were added after 30 min and 1 h,

respectively, and the reaction mixture was kept at room-

temperature overnight. Water was added and the organic

solvent was removed under reduced pressure. CH₃CN was

added to the aqueous phase and the solution was purified

immediately by preparative HPLC to yield the product.

Ben zen esu lfon ic a cid 3-[2-(4-a m in om eth yl-3-h yd r oxy-

p h en yl)eth oxy]p h en yl ester (28c): yield 60%; ¹H NMR

(CD₃OD) δ 2.99 (2H, t, J ) 6.4), 4.07-4.11 (4H, m), 6.53-6.57

(2H, m), 6.81-6.89 (3H, m), 7.18-7.26 (2H, m), 7.60-7.66 (2H,

m), 7.74-7.80 (1H, m), 7.84-7.87 (2H, m); HRMS calcd for

3,5-Dich lor o-2-m et h oxyb en zen esu lfon ic a cid 3-[2-(4-

ca r b a m im id oyl-3-h yd r oxyp h en yl)et h oxy]-5-et h ylp h en -

yl ester (41c): yield 28%; ¹H NMR (CD₃COOD) δ 1.09 (3H, t,

J ) 7.7), 2.52 (2H, q, J ) 7.7), 3.06 (2H, t, J ) 6.4), 4.06 (3H,

s), 4.15 (2H, t, J ) 6.4), 6.52-6.56 (2H, m), 6.67 (1H, s), 6.99,

1H, d, J ) 8. 1), 7.08 (1H, s), 7.65 (1H, d, J ) 2.5), 7.71 (1H,

d, J ) 8.1), 7.78 (1H, d, J ) 2.5); HRMS calcd for C₂₄H₂₄

Cl₂N₂O₆S, 538.0732; observed, 538.0727.

-

Deter m in a tion of IC₅₀of Th r om bin . A stock solution of

60 NIH U/mL human R-thrombin (cat. no. T6759, Sigma, St

Louis, MO) in buffer (0.05 M Tris-HCl, pH 7.4, I ) 0.15, 1 g/L

bovine serum albumin) was frozen at -70 °C. A working

solution was freshly prepared by dilution of the stock solution

with the assay buffer to a nominal concentration of 10 NIH

U/mL, which corresponds to an assay concentration of 0.02

NIH U/mL.

The 18 compounds were each dissolved to a concentration

of 0.1 mM in dimethyl sulfoxide (DMSO). These solutions were

then prepared to give assay concentrations in the range

between 0.06 nM and 133 µM.

C

₂₁H₂₁NO₅S, 399.1140; observed, 399.1138.

2,5-Dich lor oben zen esu lfon ic acid 3-[2-(4-am in om eth yl-

3-h yd r oxyp h en yl)eth oxy]-5-m eth ylp h en yl ester (34c):

yield 57%; ¹H NMR (CD₃SOCD₃) δ 2.21 (3H, s), 2.83 (2H, t, J

) 7.0), 3.80 (2H, s), 4.03 (2H, t, J ) 7.0), 6.42-6.43 (1H, m),

6.51-6.52 (1H, m), 6.58-6.60 (2H, m), 6.75 (1H, s), 6.95 (1H,

d, J ) 7.7), 7.90-7.91 (3H, m); HRMS calcd for C₂₂H₂₁Cl₂NO₅S,

481.0518; observed, 481.0520.

2,3-Dich lor oben zen esu lfon ic acid 3-[2-(4-am in om eth yl-

3-h yd r oxyp h en yl)eth oxy]-5-eth ylp h en yl ester (40c): yield

58%; ¹H NMR (CD₃COOD) δ 1.08 (3H, t, J ) 7.7), 2.51 (2H, q,

The chromogenic substrate pyroGlu-Pro-pNA (S-2366, Chro-

mogenix AB, Mo¨lndal, Sweden) was dissolved in water to a

3438 J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21

Linusson et al.

concentration of 10 mM and further diluted with the assay

buffer to 3.75 mM (assay concentration, 0.3 mM).

pK_avalues were calculated from electrophoretic mobility data,

analogously to the procedure described by Gluck and co-

workers.⁵⁴

The thrombin robot assays were performed in a ROBOSYS

Plato 3000 Robotic microplate processor in total volumes of

150 µM. The inhibitor (126 µL) and thrombin (12 µL) were

mixed, and the reaction was started by adding 12 µL of the

substrate working solution. The absorbance was monitored at

405 nm after 40 min of incubation. The increase in absorbance

(405 nm) was linear during the incubation time. The rate dA/

dt can therefore be approximated from ∆A 405/40 min, and K_i

can be estimated from the relation K_i) IC₅₀/(1 + [S]/K_m),

where [S] ) 0.3 mM and K_m) 0.33 mM. The compound

concentrations that inhibit 50% of thrombin (IC₅₀) activity were

determined from dose-response curves using 10 inhibitor

concentrations.

Ack n ow led gm en t. We are grateful to Dr. Angela

Menschik-Lundin for performing the biological mea-

surements, Dr. Anita Fridstro¨m for pK_ameasurements,

and Drs. J ohanna Deinum and Tord Inghardt for

valuable biological information and discussions. We are

also grateful to Dr. Karl-Erik Karlsson for recording the

HRMS spectra. Financial support from the Swedish

Natural Science Research Council (NFR) and Astra-

Zeneca R&D, Mo¨lndal, is gratefully acknowledged.

Refer en ces

Deter m in ation of Dissociation Con stan t (K_i) of Tr ypsin .

A stock solution of 1 g/L bovine trypsin (cat. no. T8003, Sigma,

St Louis, MO) in 1 mM HCl and 1 mM CaCl₂was frozen in

aliquots at -70 °C. For the steady state studies, a working

solution was freshly prepared by dilution of the stock solution

with a buffer (0.05 mol/L Tris-HCl, pH 8.2 at 20 °C, CaCl₂20

mM) to a final trypsin concentration of 0.25 mg/L.

The inhibitors were dissolved by adding 2% and 0.2%

ethanol, respectively, for compounds 29c and 32b, and by

adding 2.8% and 2.6% 2-propanol, respectively, for compounds

35c and 43b. The solutions were diluted using the trypsin

buffer to final stock concentrations of 100 µM for compounds

29c, 35c, and 43b and 10 µM for compound 32b. The stock

solutions were prepared to give seven assay concentrations of

the inhibitors (0.00, 0.01, 0.02, 0.04, 0.06, 0.08, and 0.10 µM).

The chromogenic substrate Bz-Ile-Glu-(-OR)-Gly-Arg-NH-

pNA·HCl; R ) 50% H/50% CH₃(S-2222, Chromogenix AB,

Mo¨lndal, Sweden) was dissolved in water to give three

concentrations, 0.15, 0.25, and 0.50 mM (assay concentrations,

15, 25, and 50 µM, respectively).

The measurements were carried out using a Cobas Bio

centrifugal spectroanalyzer (Roche, Basel, Switzerland) at 37

°C. The inhibitor (90 µL) and trypsin (250 µL) were incubated

for 5 min. The reaction was started by adding 40 µL of the

substrate and the change in absorbance at 405 nm of the

mixture was monitored every 10th second for 110 s. The K_i

values were calculated from the best fitting straight line

obtained by linear regression analysis.

Deter m in a tion of CE log k′ a n d p K_a. Two different

experimental descriptors were used to estimate the tendency

of the test compounds to be passively absorbed: pK_aand

migration behavior in micellar media (CE log k′). Both descrip-

tors were determined by means of capillary electrophoresis

(CE) separations. The use of micelles as separation media in

CE-MEKC mode experimentally models absorption to a lipid-

like pseudostationary phase. In this study, all 18 compounds

were subjected to MEKC separations using a sodium cholate

(SC) bile acid micellar electrolyte. From these measurements

retention factors, CE log k′ values, were calculated. pK_avalues

were measured for the compounds that contain basic residues.

Capillary electrophoresis (CE) was carried out on an HP 3D

capillary electrophoresis system (Agilent technologies, Wald-

bronn, Germany) with a diode-array detector. Untreated fused-

silica capillaries from Polymicro Technologies of 50 µm i.d.,

375 µm o.d., and 26.5 cm effective length were used for

separations. The applied voltage was 300 V/cm and the

temperature of the capillary was maintained at 37.0 ( 0.1 °C

for log k′ measurements and 22.2 ( 0.1 °C for the pK_a

measurements. Prior to each analytical separation the capil-

lary was rinsed consecutively with 0.1 M NaOH, purified

water, and electrolyte. Sample solutions were injected by

hydrodynamic injection at 20 mbar (5 s). The MEKC micellar

electrolyte was 100 mM sodium cholate (SC) in sodium

phosphate buffer (pH 7.4, I ) 0.05). Standard solutions

comprised the respective test compound and a set of marker

compounds for the measurements of CE log k′.

(1) Fenton, J . W., II Thrombin specificity. Ann. N. Y. Acad. Sci.

1981, 370, 468-495.

(2) Coughlin, S. R.; Vu, T.-K. H.; Hung, D. T.; Wheaton, V. I.

Characterization of a functional thrombin receptor. J . Clin.

Invest. 1992, 89, 351-355.

(3) Van Obberghen-Schilling, E.; Pouysse´ggur, J . Signaling path-

ways of the thrombin receptor. Thromb. Haemostasis 1993, 70,

163-167.

(4) Bajusz, S.; Barabas, E.; Tolnay, P.; Szell, E.; Bagdy, D. Inhibition

of thrombin and trypsin by tripeptide aldehydes. Int. J . Pept.

Protein Res. 1978, 123, 217-221.

(5) Kettner, C.; Shaw, E. D-Phe-Pro-Arg CH2ClsA selective affinity

label for thrombin. Thrombosis Res. 1979, 14, 969-973.

(6) Stu¨rzebecher, J .; Markwardt, F.; Voigt, B.; Wagner, G.; Wals-

mann, P. Cyclic amides of N-alpha-arylsulfonylaminocylated

4-amidinophenylalaninesTight binding inhibitors of thrombin.

Thrombosis Res. 1983, 29, 635-642.

(7) Stu¨rzebecher, J .; Walsmann, P.; Voigt, B.; Wagner, G. Inhibition

of bovine and human thrombins by derivatives of benzamidine.

Thrombosis Res. 1984, 36, 457-465.

(8) Kam, C.-M.; Fujikawa, K.; Powers, J . C. Mechanism-based

isocoumarin Inhibitors for trypsin and blood coagulation serine

proteasessNew anticoagulants. Biochemistry 1988, 27, 2547-

2557.

(9) Lee, S. L.; Alexander, R. S.; Smallwood, A.; Trievel, R.; Mers-

inger, L. et al. New inhibitors of thrombin and other trypsin-

like proteases: Hydrogen bonding of an aromatic cyano group

with a backbone amide of the P1 binding site replaces binding

of a basic side chain. Biochemistry 1997, 36, 13180-13186.

(10) Tucker, T. J .; Brady, S. F.; Lumma, W. C.; Lewis, S. D.; Gardell,

S. J . et al. Design and synthesis of a series of potent and orally

bioavailable noncovalent thrombin inhibitors that utilize non-

basic groups in the P1 position. J . Med. Chem. 1998, 41, 3210-

3219.

(11) Bachand, B.; DiMaio, J.; Siddiqui, M. A. Synthesis and structure-

activity relationship of potent bicyclic lactam thrombin inhibi-

tors. Bioorg. Med. Chem. Lett. 1999, 9, 913-918.

(12) Wiley: M. R.; Weir, L. C.; Briggs, S. L.; Chirgadze, N. Y.;

Clawson, D. et al. The design of potent, selective, noncovalent,

peptide thrombin inhibitors utilizing imidazole as a S1 binding

element. Bioorg. Med. Chem. Lett. 1999, 9, 2767-2772.

(13) Sall, D. J .; Bailey, D. L.; Bastian, J . A.; Buben, J . A.; Chirgadze,

N. Y. et al. Diamino benzo[b]thiophene derivatives as a novel

class of active site directed thrombin inhibitors. 5. Potency,

efficacy, and pharmacokinetic properties of modified C-3 side

chain derivatives. J . Med. Chem. 2000, 43, 649-663.

(14) Obst, U.; Banner, D. W.; Weber, L.; Diederich, F. Molecular

recognition at the thrombin active site: structure-based design

and synthesis of potent and selective thrombin inhibitors and

the X-ray crystal strucutre of two thrombin-inhibitor complex.

Chem. Biol. 1997, 4, 287-295.

(15) Menear, K. Progress towards the discovery of orally active

thrombin inhibitors. Curr. Med. Chem. 1998, 5, 457-468.

(16) Rewinkel, J . B. M.; Adang, A. E. P. Strategies and progress

towards the ideal orally active thrombin inhibitor. Curr. Pharm.

Des. 1999, 5, 1043-1075.

(17) Sanderson, P. E. J . Small, noncovalent serine protease inhibitors.

Med. Res. Rev. 1999, 19, 179-197.

(18) Martin, E. J .; Blaney, J . M.; Siani, M. A.; Spellmeyer, D. C.;

Wong, A. K. et al. Measuring diversity: Experimental design of

combinatorial libraries for drug discovery. J . Med. Chem. 1995,

38, 1431-1436.

(19) Lundstedt, T.; Clementi, S.; Cruciani, G.; Pastor, M.; Kettaneh,

N. et al. Intelligent Combinatorial Libraries. Computer-Assisted

Lead Finding and Optimization. Current Tools for Medicinal

Chemistry; Verlag Helvetica Chimica Acta: Basel, 1997; pp 190-

208.

Determination of pK_awas carried out using a set of 12

buffers covering a pH range from 3.1 to 12.5. The respective

(20) Linusson, A.; Wold, S.; Norden, B. Statistical molecular design

of peptoid libraries. Mol. Divers. 1999, 4, 103-114.

A Library of Thrombin Inhibitors

J ournal of Medicinal Chemistry, 2001, Vol. 44, No. 21 3439

(21) Gillet, V. J .; Willet, P.; Bradshaw, J . The effectiveness of reactant

pools for generating structurally diverse combinatorial libraries.

J . Chem. Inf. Comput. Sci. 1997, 37, 731-740.

(22) J amois, E. A.; Hassan, M.; Waldman, M. Evaluation of reagent-

based and product-based strategies in the design of combinato-

rial library subsets. J . Chem. Inf. Comput. Sci. 2000, 40, 63-

70.

(23) Linusson, A.; Gottfries, J .; Lindgren, F.; Wold, S. Statistical

molecular design of building blocks for combinatorial chemistry.

J . Med. Chem. 2000, 43, 1320-1328.

(24) Spellmeyer, D. C.; Blaney, J . M.; Martin, E. Computational

approaches to chemical libraries. Practical application of com-

puter-aided drug design; Marcel Dekker Inc.: New York, 1997;

pp 165-193.

(25) Drewry, D. H.; Young, S. S. Approaches to the design of

combinatorial libraries. Chemometrics Intell. Lab. Syst. 1999,

48, 1-20.

(26) Bode, W.; Mayr, I.; Baumann, U.; Huber, R.; Stone, S. R. et al.

The refined 1.9 Å crystal structure of human alpha-thrombin:

Interaction with D-Phe-Pro-Arg chloromethyl ketone and sig-

nificance of Tyr-Pro-Pro-Trp insertion segment. EMBO J . 1989,

8, 3467-3475.

(27) Bode, W.; Turk, D.; Karshikov, A. The refined 1.9-Å X-ray crystal

structure of D-Phe-Pro-Arg chloromethyl ketone-inhibited hu-

man alpha-thrombin: Structure analysis, overall structure,

electrostatic properties, detailed active site geometry, and

structure-function relationship. Protein Sci. 1992, 1, 426-471.

(28) Fenton, J . W., II; Ofosu, F. A.; Moon, D. G.; Maraganore, J . M.

Thrombin structure and function: why thrombin is the primary

target for antithrombotics. Blood Coagulation Fibrinolysis 1991,

2, 69-75.

(29) Olsson, T.; Sherbukhin, V. Synthesis and Structure Administra-

tion (SaSA); AstraZeneca R&D Mo¨lndal: S-431 83 Mo¨lndal,

Sweden.

(30) Andersson, P. M.; Sjo¨stro¨m, M.; Wold, S.; Lundstedt, T. Com-

parison between physicochemical and calculated molecular

descriptors. J . Chemom. 2000, 14, 629-642.

(31) Oprea, T.; Gottfries, J . Chemography, the art of navigating in

the chemical space. J . Comb. Chem. 2001, 3, 157-166.

(32) Concord; Tripos Inc.: 1699 South Hanley Rd, St. Louis, MO

63144-2913 USA.

(33) UNIX Spartan 4.0; Wavefunction, Inc.: 18401 Von Karman Ave.,

Suite 370, Irvine, CA 92612, USA.

(34) J ohnson, M. E.; Nachtsheim, C. J . Some guidlines for construct-

ing exact D-optimal designs on convex design spaces. Techno-

metrics 1983, 25, 271-277.

(35) Modde 4.0; Umetrics: Box 7960, S-907 19 Umeå, Sweden.

(36) J ackson, J . E. A user’s guide to principal components; Wiley:

New York, 1991.

(39) Stone, M. Cross-validatory choice and assessment of statistical

predictions. J . R. Stat. Soc. 1974, 36 B, 111-133.

(40) Wold, S. Cross-validatory estimation of the number of compo-

nents in factor and principal component models. Technometrics

1978, 20, 397-405.

(41) Simca-P 7.01; Umetrics: Box 7960, S-907 19 Umeå, Sweden.

(42) Trone, M. D.; Leonard, M. S.; Khaledi, M. G. Congeneric behavior

in estimations of octanol-water partition coefficients by micellar

electrokinetic chromatography. Anal. Chem. 2000, 72, 1228-

1235.

(43) Yang, S. Y.; Bumgarner, J . G.; Kruk, L. F. R.; Khaledi, M. G.

Quantitative structure-activity relationships studies with mi-

cellar electrokinetic chromatographysInfluence of surfactant

type and mixed micelles on estimation of hydrophobicity and

bioavailability. J . Chromatogr. A 1996, 721, 323-335.

(44) O¨ rnskov, E.; Gottfries, J .; Folestad, S. Correlation of drug

absorption with migration data from capillary electrophoresis

using micellar electrolytes. In preparation.

(45) Woodrow, B.; Dorsey, J . Thermodynamics of micelle-water

partitioning in micellar electrokinetic chromatographyscom-

parisons with 1-octanol-water partitioning and biopartitioning.

Environ. Sci. Technol. 1997, 31, 2812-2820.

(46) O¨ rnskov, E.; Markides, K. E.; Folestad, S. Determination of

dissociation constants of labile drug compounds by capillary

electrophoresis. Paper 160. 21st International Symposium on

Capillary Chromatography and Electrophoresis: Park City,

Utah, J une 20-24, 1999.

(47) Unpublished result; AstraZeneca R&D Mo¨lndal: S-431 83

Mo¨lndal, Sweden.

(48) Box, G. E. P.; Hunter, W. G.; Hunter, J . S. Statistics for

experimenters. An introduction to design, data analysis, and

model building; J ohn Wiley & Sons: New York, 1978.

(49) Karmarkar, S. N.; Kelkar, S. L.; Wadia, M. S. A simple unusual

one-step conversion of aromatic aldehydes into nitriles. Synth.

Comm. 1985, 510-512.

(50) ACD/ Lab; Advanced Chemistry Development Inc.: 90 Adelaide

Street West, Suite 702 Toronto, Ontario M5H 3V9, Canada.

(51) Breton, G. W. Selective monoacetylation of unsymmetrical diols

catalyzed by silica gel-supported sodium hydrogen sulfate. J .

Org. Chem. 1997, 62, 8952-8954.

(52) Original procedure: Wang, R. X.; You, X. Z.; Meng, Q. J .; Mintz,

E. A.; Bu, X. R. A modified synthesis of O-hydroxyaryl aldehydes.

Synth. Commun. 1994, 24, 1757-1760.

(53) Original procedure: Aoyama, T.; Shioiri, T. Trimethylsilyldi-

azomethane: A convenient reagent for the O-methylation of

alcohols. Tetrahedron Lett. 1990, 31, 5507-5508.

(54) Cleveland, J . A.; Benko, M. H.; Gluck, S. J .; Walbroehl, Y. M.

Automated pK_adetermination at low solute concentrations by

capillary electrophoresis. J . Chromatogr. A 1993, 652, 301-308.

(37) Ho¨skuldsson, A. PLS regression methods. J . Chemom. 1988, 2,

211-228.

(38) Wold, S. PLS for multivariate linear modeling. Chemometric

methods in molecular design; VCH: Weinheim, 1995; pp 195-

218.

J M010833F

Article Doi

Statistical molecular design, parallel synthesis, and biological evaluation of a library of thrombin inhibitors

DOI: 10.1021/jm010833f

Source and publish data:

Authors:

Article abstract of DOI:10.1021/jm010833f

Full text of DOI:10.1021/jm010833f

Products guided by the article

R&D Labs maybe for 372522-98-0

Relevant to this article

Hot Product