aspartic protease inhibitors

Eur. J. Biochem. 267, 1±9 (2000) q FEBS 2000

Aspartic protease inhibitorsAn integrated approach for the design and synthesis of diaminodiol based peptidomimetics

Alessandro Tossi1, Irena Bonin2, Nikolinka Antcheva1, Stefano Norbedo3, Fabio Benedetti3, Stanislav Miertus2,Anil C. Nair4, Tibor Maliar4, Federico Dal bello5, Giorgio PaluÁ 5 and Domenico Romeo1

1Department of Biochemistry, Biophysics and Macromolecular Chemistry, University of Trieste, Italy; 2International Centre for Science and

High Technology, UNIDO, Area Science Park, Trieste, Italy; 3Department of Chemical Sciences, University of Trieste, Italy; 4POLY-tech,

Area Science Park, Trieste, Italy; 5Institute of Microbiology, University of Padova, Italy

Aspartic proteases play key roles in a variety of pathologies, including acquired immunodeficiency syndrome.

Peptidomimetic inhibitors can act as drugs to combat these pathologies. We have developed an integrated

methodology for preparing human immunodeficiency virus (HIV)-1 aspartic protease diaminodiol inhibitors,

based on a computational method that predicts the potential inhibitory activity of the designed structures in terms

of calculated enzyme±inhibitor complexation energies. This is combined with a versatile synthetic strategy that

couples a high degree of stereochemical control in the central diaminodiol module with complete flexibility in the

choice of side chains in the core and in flanking residues. A series of 23 tetrameric, pentameric and hexameric

inhibitors, with a wide range of calculated relative complexation energies (247.2 to 1117 kJ´mol21) and

predicted hydrophobicities (logPo/w � 1.8±8.4) was thus assembled from readily available amino acids and

carboxylic acids. The IC50 values for these compounds ranged from 3.2 nm to 90 mm, allowing study of

correlations between structure and activity, and individuation of factors other than calculated complexation

energies that determine the inhibition potency. Multivariable regression analysis revealed the importance of

side-chain bulkiness and rigidity at the P2, P2 0 positions, suggesting possible improvements for the prediction

process used to select candidate structures.

Keywords: computational studies; diaminodiol inhibitors; HIV-1 protease; peptide synthesis; pseudopeptides.

Proteases play a key role in a variety of biological processes,both at the physiological level and in infection. Theirinteraction with inhibitors is the subject of intense investi-gation, also with the view to finding lead compounds for drugsthat can prevent or treat pathologies dependent on specificprotein processing or degradation [1]. One protease family thathas recently attracted much attention is that of the asparticproteases (PR) [2,3]. This family includes mammalian proteinssuch as the digestive enzyme pepsin, renin, which is involved incontrol of blood pressure, and lysosomal cathepsin D. PRs arealso key players in determining the infectiveness of variouspathogens, ranging from retroviruses to fungi and protozoa.Their inhibition thus have therapeutic value in conditionsvarying from hypertension, inflammation, tumour metastasis

and Alzheimer's disease to fungal or viral infections, andmalaria.

In the late 1980s, renin was a principal target forpeptidomimetic inhibitors with the aim of providing a selectivetherapy for hypertension [4]. Instead, the 1990s witnessed thedevelopment of hundreds of inhibitors of the PR of thehuman immunodeficiency virus (HIV) [5±8]. This intenseactivity has brought five inhibitors into clinical use in thetreatment of acquired immunodeficiency syndrome (AIDS;http://www.niaid.nih.gov/daids/dtpdb/FDADRUG.HTM). Morerecently, the candidapepsins involved in fungal invasiveness [9]and the plasmepsins involved in haemoglobin degradation bythe malarial protozoan Plasmodium falciparea [10] have alsobeen identified as potentially important therapeutic targets forPR inhibitors.

We have been developing an integrated method for thedesign and synthesis of peptidomimetic PR inhibitors based onthe modular assembly of appropriate flanking residues ontopresynthesized, nonhydrolizable central cores [11,12]. Thechoice of side-chains on the central module and flankingresidues is guided by a computational procedure that usesmolecular mechanics to calculate complexation energiesbetween the protease and candidate inhibitor structures, as ameans of predicting the inhibition potency [13±15]. Wefocused initially on a diaminodiol type core, and used HIV-1PR as the reference protease to develop our method. Thisimportant target has become the protein most studied by X-raydiffraction techniques, and numerous crystal structures withmany different types of inhibitors are thus available formodelling studies (see http://www.rcsb.org/pdb/).

Correspondence to A. Tossi, Department of Biochemistry, University of

Trieste, Via L. Giorgieri 1, I-34127 Italy. Fax: 1 39 040 676 3691,

E-mail: [email protected]

Abbreviations: Cha, cyclohexylalanyl; Chr, chromone-2-carboxylic

acid; DEcompl, calculated relative complexation energy; Dtg,

d-a-(2-thienyl)glycine; HATU, O-(7-azabenzotriazol-1-yl)-1,1,3,3-tetra-

methyluronium hexafluorophosphate; HIV, human immunodeficiency

virus; IC50, concentration for test compound for which 50% inhibition of

HIV-1 protease was observed; Kyn, kynurenic acid; logPo/w, calculated

water octanol partition coefficient; Poa, phenoxyacetic acid; PR, aspartic

protease; R, rigidity parameter; Rt, retention time; S, surface area; TBTU,

2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium tetrafluoroborate.

Enzyme: human immunodeficiency virus 1 aspartic protease

(EC 3.4.23.16).

(Received 25 October 1999, accepted 19 January 2000)

Q2

Eur J Biochem1164 EJB EB1168 aw 10/2/0 14:36 ALDEN

2 A. Tossi et al. (Eur. J. Biochem. 267) q FEBS 2000

Our synthetic approach privileges simplicity and flexibilityby fulfilling the following requirements: readily availablestarting materials; stereoselective synthesis of the centraldiaminodiol to avoid the need for diastereoisomer separation;the possibility of independently varying residues P1 and P1

0 inthe central unit; the possibility of independently varying thenumber and type of Pn, Pn

0 residues flanking the central unit oneither side. As none of the synthetic methods described in theliterature [16±22] meets all of the above requirements, wedeveloped a novel synthetic strategy that leads, in five steps, tohomochiral, Boc-monoprotected, all-S 1,4-diamino-2,3-diols,with either identical or nonidentical side chains [11,23].

Using this methodology, we prepared several pseudopeptidicinhibitors containing 4±6 residues and ranging from completelyC2 symmetrical to completely asymmetrical. Building blockswere chosen from among readily available amino acids andcarboxylic acids so as to enhance interactions with the proteaseactive site, modulate the hydrophobicity/hydrophylicity of theinhibitors and reduce the peptidic nature. The objective was toevaluate a series of compounds with sufficient moleculardiversity and range of calculated relative complexation energy(DEcompl) values to allow a quantitative correlation withinhibition activity. The IC50 values for the synthesizedcompounds were thus correlated with the DEcompl valuesusing regression analysis, allowing us to gain further insightinto factors that affect the efficiency of inhibition, and renderthe predictive process used to select designed inhibitors moreeffective. Calculated water octanol partition coefficient(logPo/w) values, as predictors of lipophilicity, were alsoconsidered to be a potentially useful parameter in the selectionprocess.

M A T E R I A L S A N D M E T H O D S

Reagents

Fmoc-protected amino acids, o-(7-azabenzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HATU) and2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium tetrafluoro-borate (TBTU) were from Perseptive Biosystems/PE (Norwalk,CT, USA). Side-chain protection for the amino acids was asfollows: g-benzyl ester for Glu, t-butyloxycarbonyl for Trp,t-butyl for Thr and Ser. Phenoxyacetic acid (Poa) andkynurenic acid (Kyn) were from Fluka Chemie (Buchs,Switzerland), chromone-2-carboxylic acid (Chr) and d-a-(2-thienyl)glycine (Dtg) were from Sigma-Aldrich (St Louis,MO, USA).

Computational method

New compounds were modelled starting from the diaminodiolicinhibitor A76928 bound to HIV-1 PR (PDB entry: 1HVK),using the builder option in the insight II program (BiosymTechnologies, San Diego, CA, USA) on a Silicon Graphicsworkstation. This crystal structure was also used as a startingstructure to calculate reference energy values. All structureswere considered to be at pH 6.5 with all the protonizable andionizable residues charged, except for the active site residuesAsp A25 and Asp B25, which shared one proton [24].Crystallographic water W415 was retained in all calculations.The geometry of the inhibitors and of protease residuesinvolved in the active site, all residues possessing atoms within3 AÊ of any atom in the inhibitor [14,25], were then optimizedby energy minimization, using the conjugate gradient method inthe discover 2.7 program (Biosym Technologies, San Diego,

CA, USA). The geometry of unbound inhibitors was optimizedseparately. To avoid hitting local minima, different startingconfigurations for both complexed and unbound structures weretested, originating from manual docking of inhibitor residues.After geometry optimization, total energies were calculated asthe sum of bonding and nonbonding (dispersion/repulsion andelectrostatic interactions, using an effective dielectric constant:� 4, to simulate the environment within the active site ofthe protein) [14,25].

The complexation energy of a reversible inhibitor (I) withHIV PR is defined by the relation:

Ecompl � E�PR:I� 2 E�PR� 2 E�I�

where E[PR:I], E[PR] and E[I] are the calculated energies ofthe enzyme±inhibitor complex, the uncomplexed enzymeand the unbound inhibitor, respectively. For structurally relatedpseudopeptides, Pn¼P1±C[X1±X2]±P1

0¼Pn', containing an[X1±X2] nonscissile bond (in this case a diol), the relativecalculated complexation energy for the designed inhibitor isgiven by DEcompl � Ecompl(I) 2 Ecompl(I*), where I and I*indicate the newly designed and reference inhibitors, respec-tively. This parameter gives an indication of how variations inthe structure of Pn residues affect binding; the more negativethe value, the better the predicted inhibition. The use of relative,rather than absolute, energies should reduce possible errorsderiving from the approximate nature of the force field andfrom neglecting entropic contributions [13,14].

Peptidomimetic synthesis

Synthesis of the mono (Boc)-protected diaminodiol and ofpeptidomimetics based on this central unit, have been describedin detail elsewhere [11,12]. Briefly, to synthesize symmetricinhibitors (PII, VI±XII, Table 1), 0.05 mmol of the diamino-diol were Boc-deprotected with 95% trifluoroacetic acid,dissolved in 100 mL N-methyl pyrrolidone and neutralizedwith DIPEA under a flow of nitrogen. To this solution wereadded 0.2 mmol each of the appropriate residue or dipeptideand TBTU or HATU, 0.34 mmol DIPEA and optionally0.005 mmol of CuCl2 to prevent racemization. Dipeptideswere presynthesized using the method described by Mitin [26].The reaction mixture was kept for 2 h at 40 8C and, ifnecessary, overnight at room temperature, monitoring thedisappearance of free diaminodiol using ES-MS. The productwas extracted from 5% KHSO4 with EtOAc, washed with brine,dried over MgSO4 and evaporated under vacuum. Fmocprotection was removed with piperidine and the productrecrystallized from EtOAc/hexane. For N-terminal acetylation,the hydroxyl groups were temporarily protected with 0.4 mmolTMSCl in 200 mL NMP at 30 8C for 2 h and the peptide treatedwith 1 mmol acetic anhydride at room temperature for 30 min.The final product was poured onto 20% citric acid, extractedand recrystallized as described above. For asymmetricinhibitors (e.g. PI, PV, PXIII±XXIII), after addition ofresidues to the unprotected side of the diol, the other sidewas deprotected with 95% trifluoroacetic acid and theappropriate residues added. For inhibitors containing Glu, afinal orthogonal removal of the Bzl side chain-protecting groupwas effected via catalytic hydrogenation. The correct structureof the inhibitors was confirmed by ES-MS and the compoundspurified by semiprepeparative C4 RP-HPLC (Waters radialpak)using a 0±60% water to acetonitrile gradient in the presenceof 0.05% trifluoroacetic acid. Overall yields after purifi-cation and molecular masses (MH1) were PI: 35%, 921.1;PII: 30%, 955.6; PIII 30%, 946.1; PIV: n.d., 943.6; PV:

Q3


q FEBS 2000 The design and synthesis of diaminodiol-based peptidomimetics (Eur. J. Biochem. 267) 3

10%, 767.8; PVI: 15%, 943.1; PVII: 10%, 893.5; PVIII:20%, 583.4; PIX: 40%, 757.3; PX: 35%, 645.1; PXI: n.d.,643.3; PXII: 55%, 569.0; PXIII: 10%, 718.5; PXIV: 30%,674.5; PXV: 40%, 760.0; PXVI: 25%, 769.6; PXVII: 35%,787.8; PXVIII: 35%, 742.5; PXIX: 20%, 744.0; PXX: 10%,743.5; PXXI: 15%, 707.0; PXXIII: 30%, 707.0; PXXIII:30%, 745.5.

Assays

IC50 values were determined using recombinant wild-typeHIV-1 PR obtained from transformed Escherichia coliHMS174(DE3)pLysS cells [27]. Assays were carried out with10 nm HIV-1 PR, using the fluorogenic substrate (S) Abz-Thr-Ile-Nle-Phe(NO2)-Gln-Arg-NH2, at 25 8C in 400 mm NaCl, 5%dimethyl sulfoxide, 100 mm Mes buffer, pH 5.5. The inhibitorRO31-8959, used as reference in this assay, had a measuredIC50 of 2 nm.

The correlation between logIC50 values and parameterspotentially involved in complexation (DEcompl, surface areasand bond rigidities) was carried out using the program datafit(v5.1, Oakdale Engineering) for multiple linear regression. Thesurface area of the key residues was determined using theBiosym Insight module, and included all side-chain atoms andCa.

Retention times were determined using a C18 analyticalcolumn (Waters Symmetry) with a 1% per min water toacetonitrile gradient in the presence of 0.05% trifluoroaceticacid. Predicted octanol/water partition constants were deter-mined with the method of Meylan & Howard [28] using thekowwin programme (v1.65, Syracuse Research Centre).

R E S U LT S A N D D I S C U S S I O N

Here, we describe an integrated approach for the computer-aided design and modular synthesis of peptidomimetic aspartic

protease inhibitors. It is based on the availability of a high-resolution crystal structure of HIV-1 PR complexed with areference inhibitor for which binding data are available. Thereference inhibitor is modified in situ to a required newstructure, and the difference in calculated complexationenergies for the new and reference structures provides anindication of the inhibitory potential of the former. CalculatedlogPo/w values were used to predict bioavailability. A series of23 symmetric or asymmetric pseudopeptides containing thediaminodiol isostere, ranging in size from four to six residueswere then assembled from a presynthesized central module andselected amino-acidic or carboxilic acid residues, usingstandard peptide synthesis techniques. The biological activityof these compounds was evaluated and compared with thecalculated DEcompl values, which ranged from 246 to1117 kJ´mol21 (Table 1). The molecular diversity and pre-dicted binding characteristics of the compounds were thussufficiently wide to permit a rigorous evaluation of theprediction methods.

Inhibitor design

Our computational approach, applied to the symmetric S,S-diaminodiol inhibitor A76928[1HVK], and its R,R-isomerA76889[1HVL], indicated that the R,R isomer binds lessstrongly (DEcompl � 117 kJ´mol21). This is consistent withpublished inhibition data [29,30] showing a significantly lowerinhibition constant for A76928 than for A76889, whereas thatfor the R,S isomer A77003 (DEcompl � 15.4 kJ´mol21) iscomparable. In the case of the diaminodiol inhibitor, Hoe/Bay793, a marked reduction in antiviral activity of the R,S-isomerwas observed in cell culture studies, even though in vitrostudies indicated a comparable enzyme inhibition activity forall three stereoisomers [31]. Thus, the S,S configuration wasselected for use in all subsequent calculations and for synthesisof the P1±C[X1±X2]±P1

0 central module.

Table 1. Experimental and calculated data for synthesized compounds.

I P3 P2 P1 -c- P10 P2

0 P30

Mr

(Da)

DEcompl

(kJ´mol21)

IC50

(nM)

log(IC50)

(min) logPow Rt

PI Ac-Phe ±Ile ±Phe ±c Phe- Glu- Phe-Ac 921 5.9 12 1.08 3.8 14

PII Ac-Trp ±Val ±Phe ±c Phe- Val- Trp-Ac 955 247.2 6 0.78 4.5 19

PIII Ac-Trp ±Val ±Phe ±c Phe- Glu- Phe-Ac 946 226.3 9 0.95 3.3 12

PIV Fmoc ±Val ±Phe ±c Phe- Val- Fmoc 943 38 4 � 103 3.60 8.4 32

PV Poa ±Val ±Phe ±c Phe- Val- Poa 767 14.2 15 1.18 5.3 20.5

PVI Chr ±Val ±Phe ±c Phe- Val- Chr 843 26.3 5 0.70 3.8 15

PVII Ac-Phe ±Ile ±Ile ±c Cha- Glu- Phe-Ac 893 100 400 2.60 4.5 17

PVIII Ac-Val ±Phe ±c Phe- Val-Ac 583 117 500 2.70 1.8 11.5

PIX Ac-Trp ±Phe ±c Phe- Trp-Ac 757 38 25 � 103 4.40 3.6 12.5

PX Chr ±Phe ±c Phe- Chr 645 42 90 � 103 4.95 2.8 13

PXI Kyn ±Phe ±c Phe- Kyn 643 38 10 � 103 4.00 5.2 17

PXII Poa ±Phe ±c Phe- Poa 569 79 2 � 103 3.30 4.3 18.5

PXIII Boc ±Phe ±c Phe- Glu- Phe-Ac 718 67 600 2.78 3.5 13.5

PXIV Boc ±Phe ±c Phe- Dtg- Poa 673 54 250 2.40 5.9 19

PXV Ac-Phe ±Glu ±Phe ±c Phe- Val-Ac 760 46 130 2.11 2.0 11.5

PXVI Ac-Trp ±Val ±Phe ±c Phe- Val-Ac 768 30.5 5 0.70 3.2 15

PXVII Ac-Trp ±Ser ±Phe ±c Phe- Kyn 787 46 4 � 103 3.60 3.5 12

PXVIII Kyn ±Val ±Phe ±c Phe- Kyn 742 1.7 4 � 103 3.60 5.1 12

PXIX Kyn ±Thr ±Phe ±c Phe- Kyn 744 218.4 2 � 103 3.30 4.1 11.5

PXX Kyn ±Val ±Phe ±c Phe- Chr 744 22.9 130 2.11 3.8 18

PXXI Kyn ±Thr ±Phe ±c Phe- Poa 707 5.9 3.2 0.50 3.6 13.5

PXXII Poa ±Thr ±Phe ±c Phe- Kyn 707 18.4 500 2.70 4.3 13

PXXIII Kyn ±Dtg ±Phe ±c Phe- Poa 745 213.8 9 0.95 5.4 17.5

Q1



Starting from this core unit, we designed either C2-symmetric or asymmetric inhibitors, with up to four additionalflanking residues P2, P2

0, P3 and P30. The synthetic methodol-

ogy developed by us allowed the freedom of independentlyvarying residues in all positions, including P1 and P1

0 (Fig. 1).Residues were selected on the basis of their potential forforming specific hydrophobic interactions and H-bondswith enzyme residues (Fig. 2). Furthermore, P2 groups(Fig. 1) were also investigated for their ability to replacelarger P2/P3 ligands, thereby reducing the molecular massof the inhibitors. Particular emphasis was placed on theP2/P2

0 position, as the dual hydrophobic/hydrophilic nature ofthe S2/S2

0 subsite (Fig. 2) permits the use of P2 residues thatcan vary considerably in their hydrophobicity and H-bondingfunctionalities.

From among the designed compounds, a subset was selectedfor synthesis (Table 1), to provide a range of DEcompl values,sizes and hydrophobicities. Compounds with a calculatedcomplexation energy comparable with or lower than thatcalculated for A76928 [29,30,32] were thus predicted to inhibitthe low nanomolar range, whereas those with more positivevalues would likely be less effective as inhibitors.

The central module

The synthetic method used to produce all-Sdiaminodiol centralmodules with high yield and purity is illustrated in Fig. 3. Ourprevious experience with reduced bond inhibitors [13] indicatedthat phenylalanine side chains were the most appropriate

substituents to place in both P1 and P10. This preference is

amply confirmed by data in the literature [6,31,33±35] andderives from the fact that the phenylalanine side chains fitnicely into the S1/S1

0-binding pockets of the enzyme. Otherhydrophobic residues in these positions might, however, alsolead to favourable interactions with HIV-1 PR, as well aspermitting a modulation of pharmacokinetic properties. Theycould also be useful in the design of inhibitors towards HIV-1mutants or other aspartic proteases. Thus, the synthetic schemehas an in-built flexibility, allowing independent variation in P1

and P10, and cyclohexylalanyl and isoleucyl side chains were

chosen to replace Phe in some inhibitors.In our approach the four-carbon skeleton of the diaminodiol

is constructed starting from a Boc-protected aminoester and analdehyde (Fig. 3), each of which can have different side chains.Functional groups are then introduced stereoselectively byreduction of the carbonyl according to Cram's rule, peracidsyn-epoxydation of the allylic alcohol and ring opening ofthe epoxide. The S-configuration of the starting aminoesterdetermines the configuration of the four adjacent stereocentersin the diaminodiol, permitting the synthesis of the requiredstereoisomer in high purity. The selective protection of only oneamino group is a key feature of this approach and allowssynthesis of both symmetric and asymmetric pseudopeptides.

Hexameric inhibitors

The structure of PI (Table 1) is based on that of a previouslysynthesized, potent reduced bond inhibitor [13]. Phe is present

Fig. 1. Side-chain structures for residues

used to construct the diaminodiol inhibitors

described in the text. The peptide backbone

is denoted by the wavy lines. Cha,

cyclohexylalanine, Dtg, d-a-(2-thienyl)glycine;

Chr, chromone-2-carboxylic acid; Kyn,

kynurenic acid; Poa, phenoxyacetic acid;

Boc, tert-butiloxycarbonyl; Ac-, acetyl; Fmoc,

9-fluorenylmethyloxycarbonyl.



in both P1 and P10; Ile in P2 forms favourable hydrophobic

contacts with the S2 subsite, whereas Glu in P20 increases the

hydrophilicity of the compound and also favours binding byinteracting electrostatically with an external shell of positivelycharged enzyme residues. For positions P3 and P3

0, Trp, Tyr orPhe were all suitable candidates and the latter was selected.This asymmetric inhibitor was readily assembled using oursynthetic approach and its IC50 value was found to be in the lownanomolar range, confirming the prediction. PII has a morehydrophobic and completely symmetric structure, based onthat of powerful diaminodiol inhibitors, such as A76928 andHOE/BAY 793, but introducing Ac-Trp into positions P3/P3

0. Itwas calculated to have a considerably negative DEcompl and isquite potent. Combining structural features from PI and PII, asin PIII, also results in a potent and quite hydrophilic inhibitor.PIV is an intermediate in the synthesis of PII and otherstructures, and can itself be considered a hexamer with the

bulky and hydrophobic Fmoc moiety in P3/P30. The positive

DEcompl, however, predicted poor binding and the IC50 value isin fact micromolar.

A successful attempt was then made to reduce the peptidicnature and molecular mass of the hexameric inhibitors, withoutdecreasing potency, by replacing the bulky acetylated Trpresidue in PII with the monocyclic phenoxyacetic acid (PV) orthe bicyclic chromone carboxylic acid (PVI). In particular, Chrcan form favourable interactions with the S3/S3

0 enzymesubsites because of hydrogen bonding between the chromonecarbonyl and Asp30, and confers hydrophilicity to the inhibitor.

Finally, PVII was designed to test the flexibility of thesynthetic method and is highly asymmetric. The substitution ofphenylalanine in P1 and P1

0 with cyclohexylalanyl (Cha) andisoleucyl side chains, however, results in a markedly positiveDEcompl value and a 2.4-fold increase in the IC50 with respect tothe Phe±Phe analogue PII.

Fig. 2. Optimized model structure of PXXIII

and schematic representation of how it

interacts with enzyme-binding subsites.

Crystallographic water molecule W415 and

enzyme catalytic aspartic residues are shown.

Fig. 3. Reaction scheme for the synthesis of homochiral, Boc-monoprotected, all-S 1,4-diamino-2,3-diol central modules used in the synthesis of

inhibitors. R1 and R1 0 substituents are as described in Fig. 1. Pn indicates up to two flanking residues. (a) 6 eq. LiCH2PO(OMe)2, THF, 278 8C, 89±92%;

(b) R10CHO, K2CO3, EtOH, 25 8C, 50±80%; (c) NaBH4, MeOH, 0 8C, 78±90%; (d) mCPBA, CH2Cl2, 25 8C, 65±70%; (e) 2 eq. Et2AlN3, toluene, 278 to

25 8C, 80%; (f) 1 atm H2, 10% Pd/C, MeOH, 100%; (g) 95% trifluoroacetic acid, . 90%; (h) 2.4 eq. Pn-COOH/TBTU/HOBt (1:1:1), DIPEA, 45 8C, 15±

55%; (i) 1.2 eq. Pn-COOH (or Pn0-COOH)/TBTU/HOBt (1:1:1), DIPEA, 45 8C, 10±40%.

Q1



The correspondence between calculated complexationenergies and log(IC50) values for the hexamers was reasonablygood (Fig. 4A), indicating that, for this type of structure,DEcompl can be used as an effective predictor of inhibitionpotency. PIV was an exception, but the very markedhydrophobicity and poor aqueous solubility probable contributeto the high IC50 values.

Tetrameric inhibitors

PVIII±PXIV (Table 1) were synthesized in an attempt to betterdefine the most suitable type of residue arrangements aroundthe core for reduced molecular mass inhibitors. Those withsymmetric structures in particular (PVIII±PXII) were easilygenerated in one-pot reactions, starting from unprotecteddiaminodiol. PVIII is a truncated version of PII, and theIC50 of this very small and hydrophilic compound remains inthe submicromolar range, confirming the validity of this corearrangement. In the PIX±PXII series, different types ofaromatic system were placed symmetrically in P2 and P2

0.The measured IC50 values for all these compounds were,however, found to be in the micromolar range, and the observedtrend does not confirm the predictions. Factors other than thoseaffecting the direct interaction of these compounds with theenzyme-binding site must, therefore, also act, as discussedbelow. Analogues of PIX and PX with Cha and Ile in P1 andP10, respectively, were also synthesized (results not shown), but

the IC50 values were comparable, indicating that the P2 and P20

side chains play the predominant role in inhibition by thesetetrameric pseudopeptides.

An alternative way of reducing the size of inhibitors is totruncate residues from only one side, resulting in structures thatwould bind in a markedly asymmetric manner. PXIII, anintermediate in the synthesis of PI and PIII, is such astructure. While considerably smaller than the lattercompounds, it maintains a submicromolar IC50. Substitutingacetylated Phe with Poa in P3 and inserting the non-proteicamino acid d-thienylglycine in P2

0 further reduced the size andpeptidic nature, as well as the IC50, in PXIV, but thehydrophobicity is increased markedly. The d-enantiomer ofthienyl glycine was selected as reversal of the thiophene sidechain projection in P2

0 favoured interaction with the S20

enzyme subsite.

Pentameric inhibitors

PXV and PXVI were obtained by trimming a terminal Trp fromPIII and PII, respectively. The latter compound, in particular,maintains good binding properties. The effect of replacingterminal Val and Trp residues was then examined. Poa in P2

0(PXXI, PXXIII) was found to be preferable to Kyn or Chr,which invariably had a negative effect on binding in thisposition (cf. PXVII, PXVIII, PXIX, PXX and PXXII), eventhough the calculated DEcompl values predict otherwise. Kynwas found to be a suitable replacement for Trp in P3, whereasThr could be used to modulate hydrophobicity in P2. Thesereplacements resulted in the low molecular mass, hydrophilicand powerful inhibitor PXXI. PXXIII, with Dtg in P2, alsoshowed good binding characteristics, but was considerablymore hydrophobic.

Correlation between DEcompl and measured IC50 values

The use of calculated DEcompl to predict the potential inhibitionefficiency of designed structures can considerably reduce the

number of candidates that need to be synthesized in order toobtain powerful lead inhibitors. However, this term onlyreflects the enzyme's affinity for a potential inhibitor structureonce it is within the binding-site of the `closed' proteinstructure. Factors such as a poor solubility (preventing theinhibitor from reaching the protease), a poor initial interactionwith the enzyme in the unbound `open' state, the inability toundergo the necessary conformational changes for binding, orthe ability to induce a conformational transition of the enzymeto the `closed' state might also play a role, however, and do notappear in this term.

Thus, while a large and positive calculated DEcompl generallycorresponds to a relatively high IC50 (Table 1), a low ornegative DEcompl does not always correspond to powerfulinhibition (Fig. 4A). Linear regression analysis was carried outin an attempt to determine what other factors could affectpotency and whether they could be included in the predictionprocess. This analysis focused on the residues present atpositions P2 and P2

0, which appear to play a particularlyimportant role in binding.

Considering the compounds present in Table 1, regressionanalysis leads to a rather poor correlation equation betweenlog(IC50) and DEcompl (all points in Fig. 4A, n � 23, r � 0.53,s � 1.16 and F � 7.8, where n, r, s and F are, respectively, thenumber of compounds considered, the correlation coefficient,the standard deviation and the significance of the regression).One possible factor underlying this unsatisfactory correlationcould be a reduced torsional flexibility due to bulky residuespresent in P2 and P2

0, as compounds with Trp, Kyn or Chr inone or both of these positions are poor binders, irrespective ofDEcompl. Furthermore, for residues such as Kyn and Chr,delocalization onto the aromatic ring structures (Fig. 1)

Fig. 4. Correlation values. (A) Correlation between measured logIC50

values for inhibitors listed in Table 1 and their calculated relative

complexation energies (DEcompl). Open circles represent bulky inhibitors

or inhibitors with `rigid' amide bonds between P2/P20 and P1/P1

0.Regression line (- -) is based on all points, while regression line (Ð) is

based only on `flexible' compounds (filled circles). (B) Correlation between

measured logIC50 and values estimated from multivariable regression

analysis including DEcompl, P2/P20 surface area (S) and rigidity (R).



might also reduce the torsional flexibility of the linkagesbetween P1/P1

0 and P2/P20. The carbonyl oxygens in these

linkages are involved in hydrogen bonding with the structuralwater molecule W415, which also forms H-bonds with residuesin the enzyme flaps [5,6] (Fig. 2). Conformational rearrange-ments leading to the formation of these important bonds may beimpaired by an increased rigidity in this region. The compoundspresent in Table 1 can thus be divided into two groups,respectively, consisting of 15 `flexible' inhibitors with smallerresidues in both P2 and P2

0 and 8 more `rigid' compounds withbulky residues in one or both positions. Considering only thefirst group (Fig. 4A, solid symbols), the correlation in factimproves considerably (n � 15, r � 0.88, s � 0.56, F � 44).

We then considered the possibility of including bulkiness and`rigidity' parameters for P2/P2

0, so that the correlation could beextended to both groups. For the first term, the surface area (S)of the side chain was used, while the rigidity parameter (R) wasassigned a value of 0, 1 or 2 depending on the number of `rigid'linkages present a the level of P2/P2

0 (e.g. 0 for PXXI, 1 forPXXII and 2 for PXI). Regression analysis leads to theequation:

log�IC50� � 20:42 1 0:06*DEcompl 1 0:01*S 1 0:84*R

and resulted in reasonable correlation (Fig. 4B; n � 23,r � 0.90, s � 0.64 and F � 26.2). The terms in this equationwere found to be not intercorrelated, indicating that bothbulkiness and bond rigidity may play a role in determiningbinding. Thus, in future, the bulkiness of residues placed into P2

or P20 and the types of linkage formed with residues in P1 and

P10 will be taken into account in designing new potential

inhibitor structures.Regression was also carried out with the further inclusion of

calculated logPo/w values, to test for a possible role ofhydrophilicity in binding. However, the correlation did notimprove (n � 23, r � 0.90, s � 0.64, F � 20) and inclusion oflogPo/w values in a three independent variable model caused adecrease in the regression significance (from F � 26.2 to 20).Analogous results were obtained using reversed-phasechromatographic retention times, rather than logPo/w, as anexperimental measure of hydrophobicity (n � 23, s � 0.66,R � 0.90, F � 18.6).

Thus, hydrophobicity does not appear to greatly influenceIC50 values, at least in the range covered by our compounds.However, it is likely to play an important role in the in vivoactivity of inhibitors, so that calculated logPo/w values can be auseful additional parameter in selecting designed structures forsynthesis. This is also in view of the fact that the inherentflexibility of our method permits considerable molecular

diversity in the design of potentially tight-binding structures,which can vary markedly in hydrophobicity (e.g. PXXI andPXXIII). We found that a reasonable relationship existsbetween logPo/w and logRt (n � 23, s � 0.72, r � 0.86,F � 59.05) (Fig. 5) indicating that the calculated partitionfunctions can, in fact, be used to predict the relative behaviourof inhibitors in an experimental interphase partitioning model.A certain scatter in the data, however, indicates that otherfactors (e.g. conformational effects) can also modulate hydro-phobicity in a manner that cannot be predicted entirely by thealgorithm used to calculate the logPo/w [36].

In conclusion, we presented an integrated method fordesigning and synthesizing potential aspartic protease inhibi-tors. This method relies on calculated complexation energiesand intrinsic structural parameters to select from in situconstructed candidate structures. The most promising candi-dates are then readily assembled from a presynthesized centralcore and suitable `off-the-shelf' flanking residues. Thisapproach privileges the complete flexibility in the choice ofside chains throughout the potential inhibitor, and is thusgenerally suitable for rapidly producing lead inhibitors topathologically important aspartic proteases, once their crystalstructures become available.

The appropriate use of flanking residues allowed us to obtainpotent inhibitors with low molecular masses and reducedpeptidic nature (e.g. PXXI and PXXIII). Correlation ofcalculated DEcompl values with measured logIC50 values forthe compounds presented in this paper indicates that factorsother than complexation energy need to be taken into accountwhen designing potential inhibitors. In particular, the flexibilityand bulkiness of residues placed in the P2 and P2

0 positionsappear critical. Furthermore, logPo/w values could be used toestimate the relative lipophilicity of the compounds, therby alsocontributing to the selection process.

A C K N O W L E D G E M E N T S

We would like to acknowledge grants from the Istitituto Superiore di SanitaÁ,

National Research Program on AIDS, from the CNR Target Project on

Biotechnology and the Italian Ministry of Universities and Research.

R E F E R E N C E S

1. Babine, R. & Bender, S. (1997) Molecular recognition of protein±

ligand complexes: applications to drug design. Chem. Rev. 97,

1359±1472.

2. Takahashi, K. (1994) Aspartic proteinases: structure, function, biology

and biomedical implications. Adv. Exp. Med. Biol. 362.

3. James, M.N. (1998) Aspartic proteinases: retroviral and cellular

enzymes. Adv. Exp. Med. Biol. 436.

4. Hoover, D.J., Lefker, B.A., Rosati, R.L., Wester, R.T., Kleinman, E.F.,

Bindra, J.S., Holt, W.F., Murphy, W.R., Mangiapane, M.L., Hockel,

G.M., Williams, I.H., Smith, W.H., Gumkowski, M.J., Shepard, R.M.,

Gardner, M.J. & Nocerini, M.R. (1995) Discovery of inhibitors of

human renin with high oral bioavailability. Adv. Exp. Med. Biol. 362,

167±180.

5. West, M.L. & Fairlie, D.P. (1995) Targeting HIV-1 protease: a test of

drug-design methodologies. Trends Pharm. Sci. 16, 67±75.

6. Kempf, D.J. & Sham, H.L. (1996) HIV protease inhibitors. Curr.

Pharmac. Design 2, 225±246.

7. Martin, J.A., Redshaw, S. & Thomas, G.J. (1995) Inhibitors of HIV

proteinase. Prog. Med. Chem. 32, 239±287.

8. Moyle, G. & Gazzard, D. (1996) Current knowledge and future

prospects for the use of HIV-1 protease inhibitors. Drugs 51,

701±712.

9. Goldman, R.C., Frost, D.J., Capobianco, J.O., Kadam, S., Rasmussen,

Fig. 5. Correlation between calculated octanol/water partition coeffi-

cients (logPo/w) and experimental RP-HPLC retention times for

compounds described in this work.

Q1

Q4

Q5



R.R. & Abad-Zapatero, C. (1995) Antifungal drug targets: Candida

secreted aspartyl protease and fungal wall beta-glucan synthesis.

Infect. Agents Dis. 4, 228±247.

10. Rosenthal, P.J. (1998) Proteases of malaria parasites: new targets for

chemotherapy. Emerg. Infect. Dis. 4, 49±57.

11. Benedetti, F., Miertus, S., Norbedo, S., Tossi, A. & Zlatoidsky, P.

(1997) Versatile and stereoselective synthesis of diamino diol

dipeptide isosteres, core units of pseudopeptide HIV protease

inhibitors. J. Org. Chem. 62, 9348±9353.

12. Tossi, A., Antcheva, N., Benedetti, F., Norbedo, S., Miertus, S. &

Romeo, D. (1999) Flexible synthesis of symmetric and non-

symmetric HIV-1 protease inhibitors based on all-S-diaminodiol

isosteres. Peptide Protein Lett. 6, 145±148.

13. Tossi, A., Antcheva, N., Romeo, D. & Miertus, S. (1995) Development

pseudopeptide inhibitors of HIV-1 aspartic protease: analysis and

tuning of the subsite specificity. Peptide Res. 8, 2±8.

14. Miertus, S., Furlan, M., Tossi, A. & Romeo, D. (1996) Design of new

inhibitors of HIV-1 aspartic protease. Chem. Phys. 204, 173±180.

15. Nair, C.A., Miertus, S., Tossi, A. & Romeo, D. (1998) A computational

study of the resistance of HIV-1 aspartic protease to the inhibitors

ABT-538 and VX-478 and design of new analogues. Biochem.

Biophys. Res. Commun. 242, 545±551.

16. Kempf, D.J., Norbeck, D.W., Codacovi, L., Wang, X.C., Kohlbrenner,

W.E., Wideburg, N.E., Paul, D.A., Knigge, M.F., Vasavanonda, S.,

Craig-Kennard, A., Saldivar, A., Rosenbrook, W., Clement, J.J.,

Plattner, J.J. & Erickson, J. (1990) Structure-based C2 symmetric

inhibitors of HIV-1 protease. J. Med. Chem. 33, 2687±2689.

17. Chenera, B., Boehm, J.C. & Dreyer, G.B. (1991) Synthesis of C-2-

symmetric and pseudosymmetric HIV-1 protease inhibitors from

d-mannitol and d-arabitol. Bioorg. Med. Chem. Lett. 1, 219±222.

18. Ghosh, A.K., McKee, S.P. & Thompson, W.J. (1991) Stereocontrolled

synthesis of HIV-1 protease inhibitors with C2 axis of symmetry.

Tetrahedron Lett. 32, 5729±5732.

19. Kempf, D.J., Sowin, T.J., Doherty, E.M., Hannick, S.M., Codacovi,

L., Henry, R.F., Green, B.E., Spanton, S.G. & Norbeck, D.W.

(1992) Stereo controlled synthesis of C2-symmetric and pseudo

C2-symmetric diamino alcohols and diols for use in HIV protease

inhibitors. J. Org. Chem. 57, 5692±5700.

20. Konradi, A.W. & Pedersen, S.F. (1992) Pinacol homocoupling of (S)-2-

[N-(benzyloxy carbonyl) amino] aldeheydes by [V2Cl3(THF)6]2

[Zn2Cl6]. Synthesis of C2-symmetric (1S,2R,3R,4S)-1,4-diamino-

2,3-diols. J. Org. Chem. 57, 28±32.

21. Kang, S.H. & Ryu, D.H. (1996) Double asymmetric iodamination;

synthesis of C2 symmetric meso-amino alcohol. J. Chem. Soc. Chem.

Commun. 355±356.

22. Gurjar, M.K., Pal, S., Rao, A.V.R., Pariza, R.J. & Chorgade, M.S.

(1997) Synthesis of novel C2-symmetric and pseudo C2-symmetric

based diols, epoxides and dideoxy derivatives of HIV protease.

Tetrahedron 53, 4769±4778.

23. Benedetti, F., Berti, F. & Norbedo, S. (1998) Regio- and stereoselective

ring opening of 2,3-epoxyalcohols with diethylaluminium azide.

Tetrahedron Lett. 39, 7971±7974.

24. Wang, Y.X., Freedberg, D.I., Yarnazki, T., Wingfield, P.T., Stahl, S.J.,

Kaufman, J.D., Kiso, Y. & Torchia, D.A. (1996) Solution NMR

evidence that the HIV-1 protease catalytic aspartyl groups have

different ionization states in the complex formed with the asymmetric

drug KNI-272. Biochemistry 35, 9945±9950.

25. Miertus, S., Antcheva, N., Tossi, A., Bizik, F., Benedetti, F., Gennaro,

R. & Romeo, D. (1995) CADD of new pseudopeptide inhibitors of

HIV-1 aspartic protease. In QSAR and Molecular Modelling:

Concepts, Computational Tools and Biological Applications (Sanz,

F., Giraldo J. & Manaut, F., eds.), pp. 567±572. Prous Science

Publishers, Barcelona, Spain.

26. Mitin, Y. (1996) An effective organic solvent system for the dissolution

of amino acids. Int. J. Peptide Protein Res. 48, 374±376.

27. Maschera, B., Furfine, E. & Blair, E.D. (1995) Analysis of resistance to

human immunodeficiency virus type 1 protease inhibitors by using

matched bacterial expression and proviral infection vectors. J. Virol.

69, 5431±5436.

28. Meylan, W.H. & Howard, P.H. (1995) Atom/fragment contribution

method for estimating octanol/water partition coefficients. J. Pharm.

Sci. 84, 83±92.

29. Kempf, D.J. (1994) Design of symmetry-based, peptidomimetic

inhibitors of human immunodeficiency virus protease. Methods

Enzymol. 241, 334±354.

30. Hosur, M.V., Bhat, T.N., Kempf, D.J., Baldwin, E.T., Liu, B., Gulnik,

S., Wideburg, N.E., Norbeck, D.W., Appelt, K. & Erickson, J.W.

(1994) Influence of stereochemistry on activity for C2 symmetry-

based diol inhibitors of HIV-1 protease. J. Am. Chem. Soc. 116,

847±855.

31. Budt, K.-H., Peyman, A., Hansen, J., Knolle, J., Meichsner, C.,

Paessens, A., Ruppert, D. & Stowasser, B. (1995) HIV-1 protease

inhibitor HOE/BAY 793, structure±activity relationships in a series

of C-2-symmetric diols. Bioorg. Med. Chem. 3, 559±571.

32. Kempf, D.J., Codacovi, L., Wang, X.C., Kohlbrenner, W.E., Wideburg,

N.E., Saldivar, A., Vasavanonda, S., Marsh, K.C., Bryan, P., Sham,

H.L., Green, B.E., Betebenner, D.A., Erickson, J. & Norbeck, T.W.

(1993) Symmetry-based inhibitors of HIV protease. Structure±

activity studies of acylated 2,4-diamino-1,5-diphenyl-3-hydroxy-

pentane and 2,5-diamino-1,6-diphenylhexane. J. Med. Chem. 36,

320±330.

33. Dreyer, G.B., Boehm, J.C., Chenera, B., DesJarlais, R.L., Hassell,

A.M., Meek, T.D. & Tomaszek, T.A. (1993) A symmetric inhibitor

binds HIV-1 protease asymmetrically. Biochemistry 32, 937±947.

34. Dunn, B., Gustchina, A., Vlodawer, A. & Kay, J. (1994) Subsite

preferences of retroviral proteinases. Methods Enzymol. 241,

254±278.

35. Lange-Savage, G., Berchtold, H., Liesum, A., Budt, K.H., Peyman, A.,

Knolle, J., Sedlacek, J., Fabry, M. & Hilgenfeld, R. (1997) Structure

of HOE/BAY 793 complexed to human immunodeficiency virus

(HIV-1) protease in two different crystal forms. Structure/function

relationship and influence of crystal packing. Eur. J. Biochem. 248,

313±322.

36. Karajiannis, H. & van de Waterbeemd, H. (1995) The prediction of the

lipophilicity of peptidomimetics. A comparison between experi-

mental and theoretical lipophilicity values of renin inhibitors and

their building blocks. Pharmaceutica Acta Helvetiae 70, 67±77.

Q6



Author Queries

Q1 Au: The design and synthesis of diaminodiol-basedpeptidomimetics - running head too long. Pleasesupply an alternative of not more than 50 characters(inc. spaces)

Q2 Au: 247.2 to 1117 kJ´mol21. kJ and not kCal isused in all EJB papers. All values have beenconverted, please check carefully.

Q3 Au: Please define: TMSCl

Q4 Au: Please supply the page range

Q5 Au: Please supply the page range

Q6 Au: Any volume number

Q1


aspartic protease inhibitors

Documents