ACTAUNIVERSITATISUPSALIENSISUPPSALA2007

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 270

Molecular Simulation of EnzymeCatalysis and Inhibition

SINISA BJELIC

ISSN 1651-6214ISBN 978-91-554-6794-6urn:nbn:se:uu:diva-7468

Papers included in the thesis

I. Ersmark, K., Feierberg, I., Bjelic, S., Hulten, J., Samuelsson, B., Åqvist, J., and Hallberg, A. (2003). C2-symmetric inhibitors of Plas-modium falciparum plasmepsin II: Synthesis and theoretical predic-tions, Bioorg. Med. Chem. 11, 3723-3733.

II. Ersmark, K., Feierberg, I., Bjelic, S., Hamelink, E., Hackett, F., Blackman, M. J., Hulten, J., Samuelsson, B., Åqvist, J., and Hall-berg, A. (2004). Potent inhibitors of the Plasmodium falciparum en-zymes plasmepsin I and II devoid of cathepsin D inhibitory activity, J. Med. Chem. 47, 110-122.

III. Bjelic, S., and Åqvist, J. (2004). Computational prediction of struc-ture, substrate binding mode, mechanism, and rate for a malaria pro-tease with a novel type of active site, Biochemistry 43, 14521-14528.

IV. Bjelic, S., and Åqvist, J. (2006). Catalysis and linear free energy re-lationships in aspartic proteases, Biochemistry 45, 7709-7723.

V. Bjelic, S., Nervall, M., Gutiérrez-de-Terán, H., Ersmark, K., Hall-berg, A., and Åqvist, J. Computational inhibitor design against ma-laria plasmepsins. Manuscript.

VI. Bjelic, S., Brandsdal, B.O., and Åqvist, J. Cold and heat adaptation of citrate synthase: Effects on the general base catalyzed keto-enol isomerization step. Manuscript.

Related publications

I. Åqvist, J., Wennerstrom, P., Nervall, M., Bjelic, S., and Brandsdal, B. O. (2004). Molecular dynamics simulations of water and biomolecules with a Monte Carlo constant pressure algorithm, Chem. Phys. Lett.384, 288-294.

CONTENTS

MALARIA ........................................................................................................9Introduction ................................................................................................9Plasmepsin II ............................................................................................13

Plm II peptide cleavage (paper IV)......................................................13Plm II inhibitor development (papers I and II) ...................................20

Histo-aspartic protease (paper III)............................................................26

COLD/HEAT ADAPTATION ......................................................................28Introduction ..............................................................................................28Catalysis in temperature adapted citrate synthases (paper VI).................29

EPILOGUE ....................................................................................................32

THEORETICAL METHODS .........................................................................36Force fields...............................................................................................36

Solvent .................................................................................................37Docking ....................................................................................................37Scoring .....................................................................................................38Molecular Dynamics ................................................................................39Statistical mechanics ................................................................................40Free energy perturbation (FEP)................................................................41Linear interaction energy (LIE) method...................................................42Empirical valence bond (EVB) method ...................................................43

SUMMARY IN SWEDISH .............................................................................45Malaria .....................................................................................................45Temperatur-optimering av citratsyntas ....................................................47

ACKNOWLEDGMENTS .............................................................................49

REFERENCES ................................................................................................51

ABBREVIATIONS

Cat Cathepsin

EVB Empirical Valence Bond

FEP Free Energy Perturbation

FF Force Field

HAP Histo-Aspartic Protease

HIVP HIV-1 protease

LIE Linear Interaction Energy

MD Molecular Dynamics

MM Molecular Mechanics

Plm Plasmepsin

QM Quantum Mechanics

9

MALARIA

IntroductionDrug resistance, complex life cycle and disease spreading through a mos-quito vector are three major reasons behind malaria being such a burden to humanity. Malaria is mostly present in Africa, Asia and South America and infects hundreds of millions each year. Several species of malaria are known to infect human: Plasmodium falciparum, Plasmodium malariae, Plasmo-dium ovale and Plasmodium vivax. Falciparum malaria is most dangerous with a death rate of two to three millions a year, making it one of the deadli-est diseases in the world. [1-7]

Malaria is transmitted from human to human by female mosquitos. In Africa, the risk of acquiring malaria is especially high, primarily due to the long life length of the mosquito Anopheles gambiae. The species are endemic to the continent and are able to bite humans on average 200 times more during their life span compared to other mosquitoes [2]. When biting a malaria in-fected host, the parasite is transferred to a mosquito where it undergoes a transformation. The parasite, taken up with blood, is characterized by sexu-ally mature gametocytes that go through a reproduction phase in the gut of the mosquito. From the gut the next stage in the life-cycle, represented by sporozoites, continues in the salivary glands. From the salivary glands ma-laria parasites are injected during feeding into human hosts. [8]

In humans, sporozoites travel to the liver where they invade liver cells and through an asexual transformation generate merozoites. Unfortunately, the liver cells are only a stage-post for the second invasion that befalls red blood cells.[9] In the red blood cells merozoites multiply going through an addi-tional trophozoite form. The invasion and rapture of the red blood cells leads eventually to ten percent of their total population being infected. Sometimes during this cycle another form of the parasites emerges. These are the game-tocytes that are taken up by the mosquito for the next round in the spreading of the infection.

10

The complex life cycle of the malaria parasite gives several possible alterna-tives for the control of the infection. Mosquito eradication through insecti-cides and habitat transformation is one way of preventing the disease spread-ing. This is probably the most desirable method, but it is extremely difficult and inefficient due to the large areas involved. Alternatively, the human host has to be immunized against the malaria parasite through vaccines. Several possible vaccine targets have been under development. They are focused mainly on the pre-erythrocytic life stage, but also blocking the transmission to the mosquito is achieving attention. Although some projects are especially promising a functioning vaccine is at least a decade in future. [4]

Existing drugs target mainly the blood-stage of malaria parasite hindering the asexual reproduction in red blood cells. The most known drugs are based on quinine that disrupts the hemoglobin metabolism of the malaria parasite, leading eventually to starvation. Antifolate drugs, another large group of antimalarials, inhibit the function of the folate metabolism and are often used in combination with quinine derivatives. During the last decades a new type of drugs based on artemisinin scaffold have emerged. These drugs and their derivatives have been found to effectively reduce parasitemia mainly by inhibiting calcium dependent ATPase. Antibiotics have also been found to exhibit antimalarial activity. The inhibition mechanism is characterized by the fact that the parasite possesses an organelle, apicoplast, with prokaryotic-like functions. Since the apicoplast is responsible for the self-replication the antibiotics target for example protein synthesis. [1, 10-12]

Unfortunately drug resistance is diminishing the effectiveness of the existing malaria drugs. The malaria parasite is evolving new ways of escaping the drugs by random mutations in the genetic code. For example a single muta-tion, Ser108Asn, in dihydrofolate reductase confers resistance to antifolate drugs. [13] The existing drugs become less effective against their targets and finally do not have any effect against some strains. Combination of several drugs may still work, but in the end there is a pressing need for new and more effective therapies. The genome sequencing project of Plasmodium falciparum has revealed new potential drug targets for the inhibitor devel-opment research. Totally it was predicted that the number of proteins in the malarial genome is approximately 5000, but not all of them have a deter-mined function and about 60 % are hypothetical proteins (in year 2002). By comparing human and malaria genomes, the differences can be exploited. For instance, metabolic pathways and different transcriptional and transla-tional mechanisms specific for Plasmodium species might be used in new therapies. Specific channels and transporters can be additionally exploited because they are important not only as a drug targets but also in the delivery of drugs. This has led to the development of new potential inhibitors that are either derivatives of the existing drugs, e.g. artemisinins. Alternatvely they are targeting heme crystallization, proteases involved in hemoglobin degra-

11

dation, fatty acid metabolism that is specific for malaria parasite, choline uptake essential for membrane synthesis, glycolysis and fernesyl transferase involved in signal transduction. [1, 12, 14-16]

Figure 1: Hemoglobin is degraded in the malaria parasite’s food vacuole by plasmepsins, falcipains, falcilysins and dipeptidyl aminopeptidase 1 (DPAP1) [17].

The research presented in this thesis is mainly focused on enzymes involved in the hemoglobin degradation. During the intra-erythrocytic life stage ma-laria utilizes host hemoglobin as a food source which is initially transported to the parasite’s food vacuole where the degradation takes place. There are two protease families responsible for the hemoglobin degradation: aspartic proteases and cysteine proteases called plasmepsins and falcipains, respec-tively (Figure 1). In total, there are ten different plasmepsin genes reported in the Plasmodium falciparum genome, but only four of these are localized in the food vacuole and active during different stages of hemoglobin break-down: Plasmepsin I (Plm I), Plm II, Plm IV and histo-aspartic protease (HAP). Among falcipains, falcipain-2 (falcipain-2 ) and falcipain-3 are in-volved in the hemoglobin degradation pathway while falcipain-1 is mainly used for erythrocyte invasion. Falcipains have also been suggested to be involved in the plasmepsin processing and activation. Altogether these en-zymes are essential for the survival of the malaria parasite and are potential drug targets in the development of new therapies. [18] This thesis concerns work that is related to the P. falciparum Plm I, Plm II and HAP enzymes.

12

Figure 2: Plasmepsin II (1LF4 [19]) is an aspartic protease with two aspartic acids (in yellow) situated in the cleft between the domains. Two views are presented from top (A) and from side (B). Water molecule situated between the aspartic acids is represented by red sphere.

13

Plasmepsin II

Plm II peptide cleavage (paper IV)Plasmepsin II is an aspartic protease with two aspartic acids situated in the catalytic cleft (Figure 2) [19]. During the reaction one of the aspartic acids is negatively charged while the second one is neutral. The negatively charged aspartate abstracts a proton from a water molecule that attacks the scissile bond carbonyl carbon atom (Figure 3). The transiently formed tetrahedral intermediate (TI) dissociates when the general base aspartate functions in-stead as a general acid and protonates the nitrogen of the peptide bond. [20-

29]Alternative mechanisms have been also proposed which are less plausible:

Covalently attached reaction intermediate.This mechanism is similar to serine type proteases and has been ini-tially proposed but is now ruled out [22]. The reaction would proceed through tetrahedral intermediate that is formed during the nucleo-philic attack.

O-protonated amide [30, 31].The peptide dissociation proceeds at low pH through protonation of the amide oxygen. If the same mechanism occurs in aspartic prote-ases this would leave both aspartates negatively charged in close proximity to each other. [32-34]

Doubly protonated catalytic aspartic acids [35].The water molecule situated between the catalytic aspartates is de-protonated by the outer oxygens of these aspartic residues. This mechanism seems to have the negatively charged tetrahedral inter-mediate without any stabilization [22].

Plasmodium falciparum Plm II has been most extensively characterized among the plasmepsins with several X-ray structures reported in the PDB (1LEE, 1LF2, 1LF3, 1SME and 2BJU in Table 1) [19, 36-39]. These are co-crystallized with several different hydroxyethylamine/statine based inhibi-tors that are transition state mimetics as well as an achiral inhibitor (Figure 4) [19, 36, 38]. The structures of uncomplexed plasmepsin (1LF4) [19] and pro-plasmepsin (1PFZ) have also been determined [37], and seven additional structures are deposited in the PDB and are awaiting publication in near fu-ture (1M43, 1ME6, 1XDH, 1XE5, 1XE6, 1W6H and 1W6I).

14

Figure 3: Reaction mechanism catalyzed by aspartic proteases (Plm II, Cat D and HIVP) and HAP proceeds through a tetrahedral intermediate (TI) state. [20-23, 26-29] The formation and dissociation of the tetrahedral intermediate is either stepwise (legs of the triangle) or concerted (hypotenuse). R and P are reactant and product states, respectively.

In paper IV several question regarding aspartic proteases plasmepsin II, cathepsin D (Cat D) and HIV protease (HIVP) were addressed to learn more about their catalytic mechanism:

What is the origin of catalytic effect?

What specific interactions are important for binding and stabilization

of the tetrahedral intermediate?

Do these enzymes obey a linear free energy relationship (LFER)?

15

Table 1: Fourteen plasmepsin II structures are deposited in the PDB with chemically diverse ligands, indicating the importance of plasmepsin inhibitor development.

STRUCTURE LIGAND1LEE [36] rs367 1LF2 [36] rs370 1LF3 [19] EH58 1LF4 [19] proplasmepsin 1M43 [40] pepstatine A 1ME6 [40] statine-based 1PFZ [37] proplasmepsin 1SME [39] pepstatine A 1W6H [40] pepstatine analogue 1W6I [40] pepstatine A 1XDH [40] pepstatine A 1XE5 [40] pepstatine analogue 1XE6 [40] pepstatine analogue 2BJU [38] achiral

To investigate these questions, experimentally measured reaction rates have to be compared to calculated free energies. The transition state theory (TST) relates the chemical reaction rate, k, to the activation free energy, GTS, by

expTS

Bk T Gkh RT

(1)

where , kB, T and h are the transmission coefficient, the Boltzmann con-stant, the temperature, and the Planck’s constant, respectively [41-43]. =1 is assumed in the classical TST theory and is used throughout the theses [41-43].It reflects the dynamical property that, after reaching the transition state, the reaction proceeds to products without re-crossing back to reactants.

The activation free energies in paper IV were determined using the empirical valence bond (EVB) method together with free energy perturbation (FEP) calculations. The system was propagated through time with molecular dy-namics (MD) simulations. The ability to generate reliable free energy pro-files is the major advantage of the EVB/FEP/MD method. The reaction rates in enzymes are determined after the water reaction is parameterized to re-produce the relevant data for peptide hydrolysis in solution. The same pa-rameters are then used unchanged in the enzyme simulations, and in our case the activation barriers were calculated for six amino acid long substrates in plasmepsin II, cathepsin D and HIV protease. The calculated activation bar-

16

riers were approximately within 2 kcal/mol from the observed values. Thus the associated errors are small relative to the reduction of the enzymatic ac-tivation barrier relative to the water reaction, which is higher by 13 kcal/mol.

Figure 4: Inhibitors present in the X-ray structures are predominantly based on pepstatin A, a very potent transition state analogue. PDB codes are in the parentheses, and corresponding references are found in Table 1.

The established free energy profiles (cf. Fig.1 in paper IV) make it possible to determine the origin of catalysis by investigating the energetics associated with the transition state stabilization in enzyme and water, respectively. The results point towards the electrostatic stabilization and the preorganized ac-tive site as the major catalytic effects that enable Plm II to accelerate the reaction relative water. The most important contributions to the electrostatic stabilization during catalysis were provided by interactions with the amino acids in the active site. The electrostatics are separated and compared indi-vidually by investigating residue interactions with the catalytic region at the tetrahedral intermediate state. The residues in proximity of the reaction cen-tre are naturally having the most pronounced effect on the reaction rate, e.g.Asp34, Gly36, Tyr77, Gly216, Thr217 and Tyr192 (Figure 5 and Fig.7 in paper IV).

17

Figure 5: Plasmepsin II (green) stabilizes the tetrahedral intermediate (yel-low) by interactions from Asp34, Gly36, Tyr77, Gly216, Thr217 and Tyr192.

18

The catalytic effect on peptide cleavage was investigated additionally in two other aspartic proteases, Cat D and HIVP. Cat D is involved in human im-mune response, and as such new drugs have to be selective against it. HIVP is in contrast essential for the maturation of HIV virus particles. HIVP is included in paper IV as an interesting test case and vast amount of data that has been accumulated about its function and mechanism. Despite Plm II/Cat D and HIVP being both aspartic proteases there are some structural differ-ences which are interesting from the catalysis viewpoint [22, 44]:

HIVP is a dimer with two flaps covering the active site. Between these flaps and the substrate a water molecule is situated. Plm II and Cat D have only one flap and the active site is more open to water [22,

44].

The consensus sequence for aspartic proteases, Asp-Thr/Ser-Gly, continues with alanine in HIVP but is serine or threonine in Plm II/Cat D [22, 44].

The Cat D and HIVP reaction free energy profiles were calculated for pep-tide substrates IAFFSR and KILFLD/QVLAIA, respectively. Overall, gen-eral agreement for Cat D and HIVP peptide cleavage was found with Plm II. It is particularly interesting to try to explain the mutation of serine or threonine residues in Plm II/Cat D to alanines in HIVP. These residues inter-act with the outer oxygens of the catalytically important aspartic acid resi-dues and are significant for stabilization of the negative charge. The strong interaction of the reacting region with Gly27 in HIVP partially compensates the alanine mutation. Also the different pH sensitivity of Plm II, Cat D, and HIVP activity is of importance. HIVP has a pH optimum at 6 while it is 3 and 5 for Plm II and Cat D, respectively. The additional hydrogen bond to the general base/acid aspartate, in the active site of Plm II/Cat D, is indispen-sable because it protects the charge at low pH.

Part of the electrostatic stabilization of the reacting fragments in HIVP is provided by water molecules. The water positioned between the flaps and the substrate contributed as much as 4 kcal/mol to the interaction with the reacting groups in the tetrahedral intermediate state. Four additional water molecules positioned at the substrate ends interacted similarly yielding a total of 8 kcal/mol. In Plm II the water molecules are more of structural im-portance, bridging the interactions between the residues in close vicinity to the active site. For example the water molecule positioned between Tyr77, Ser37 and Asn39 in Plm II was predicted by MD simulations. This water is conserved across pepsin-like aspartic proteases and has been proposed to be important for the reaction mechanism [45]. Consequently, the structurally conserved water molecules in the HIVP and Plm II/Cat D proteases are

19

highly important structurally and energetically (Fig. 4 and Fig. 5 in paper IV).

Figure 6: Free energy profile for water and plasmepsin II peptide bond cleav-age. Catalysis of RMFLSF peptide in the enzyme reaction is evident com-pared to water with 15 kcal/mol stabilization.

In addition to electrostatics the reorganization energy also makes a signifi-cant contribution to catalysis. When the reaction occurs in water solution water dipoles are following the charge transfer and reorientating to interact favourably. In enzymes, the dipole moments, arising from amino acid side chains and protein backbone, are often already orientated favourably to sta-bilize the transition state. Differences in this so-called reorganization free energy provide the additional stabilization during the reaction. The reorgani-zation energy, , is related to the activation and reaction free energies by simple Marcus relationship

20

4TS

GG (2)

where G0 is the difference in the free energy between the reactants and products. Although this case holds only for diabatic reactions and is a rela-tively simplified description of the present reaction, it can still be used to calculate and compare the reorganization energies in enzyme and water reac-tions. Thus by using the calculated activation and reaction free energies for

20

the TI formation (from Table 2 in paper IV) in eq. 2, the reorganization en-ergy for water and aspartic proteases can be estimated to 37 kcal/mol and 67 kcal/mol, respectively. The simulated mechanism for the aspartic protease enzymes Plm II, Cat D and HIVP was also supported by linear free energy relationships (LFERs). LFERs are a specific case of the Marcus relationship which behaves linearly within reasonable limits of the of reaction free en-ergy. Trends are easily discerned with LFERs, as well as deviations from them allowing a systematic study of enzymatic activity. [46-51] That reactions in water obey LFERs is a fact long known [52], but also enzymes can follow such simple relationships [46-51]. The calculated values for the activation and reaction free energy of tetrahedral intermediate formation fall on the same line (Fig. 10 in IV), while the water reaction is offset by several kcal/mol. This is a manifestation of the fact that the reorganization energy is higher in water than in the enzymes.

Moreover, in paper IV the free energy profiles were evaluated for the whole reaction in Plm II with the RMFLSF (Figure 6) peptide substrate and for HIVP KILFLD. The reaction barriers were calculated to determine the high-est point along the reaction coordinate for comparison to several studies, where the rate-limiting step was determined. For example, the isotope effect studies have put forward a hypothesis where the rate-limiting step was the tetrahedral intermediate dissociation [24, 25, 31]. From the free energy profiles in paper IV the activation barrier for the TI formation and dissociation in HIVP and Plm II were of similar height, with tendencies towards the TI for-mation being the rate limiting step (Fig.1 in IV).

Molecular simulations (EVB/FEP/MD) have thus been used in paper IV to determine the catalytic effect of Plm II, Cat D and HIVP. Despite large structural differences, these aspartic proteases have been found to similarly cleave peptides very efficiently with the combination of electrostatics and preorganized active sites being the main source of transition state stabiliza-tion.

Plm II inhibitor development (papers I and II)Enzyme interactions are optimized for stabilization of the transition state during catalysis. These interactions are often used in inhibitor design, which is made to resemble the transition state instead of the substrate [53-55]. This is best understood from the thermodynamic cycle of Figure 7. The enzyme/transition-state complex is hypothetically reached either by first generating the transition state in water that binds to the enzyme in the next step, or by substrate binding followed by transition state generation in the enzyme.

21

Figure 7: The binding free energy of the transition state, G‡bind, is, compared

to the free energy of substrate binding, Gbind, enhanced by the rate accelera-tion during catalysis ( G‡

enz G‡wat), as demonstrated by the thermodynamic

cycle. The substrate (sub) is in our case, for example, the RMFLSF peptide cleaved by Plm II, and E is the enzyme.

This leads to the difference of the activation free energy in the enzyme and water being utilized in binding of the transition state:

‡ ‡ ‡bind bind enz watG G G G (3)

Thus inhibitors that are designed to mimic the transition state, also called transition state analogues, are often a first step in the rational development of potential drugs. The 1,2-dihydroxyethylene inhibitors towards plasmepsin II, presented in paper I and paper II, are similarly related to the tetrahedral transition state occurring during peptide hydrolysis (Figure 8). The transition state is represented by hydroxyls that are able to interact with the catalytic aspartates. The amino acid sidechains on the either side of the scissile bond are replaced by the vinyl sidechains. In addition, the methionine and serine side chains in the P2 and P2 positions, respectively, are substituted with the hydrophobic valine residues. The binding free energy, for this inhibitor and several additional compounds based on the same scaffold, was determined by computational techniques as well as by experimental assays (Figure 9).Molecular dynamics (MD) simulations in the combination with the linear interaction energy (LIE) method were used to theoretically predict binding free energies (cf. Theoretical Methods section for a detailed presentation of the LIE method). These were in good agreement with the experimental

22

values which makes it possible to relate them to the enzyme inhibitorcomplex structure.

Figure 8: Inhibitor based on the dihydroxyethylene scaffold (left) resembles the transition state during the reaction (right). P is the position of the amino acid in the peptide, with the prime used after the scissile bond in the N-terminal to C-terminal direction.

The 1,2-dihydroxyethylene compounds were synthesized by allylating or benzylating bislactones, and subsequently opening the rings in presence of either D- or L-valine. Only the SRRRRS stereoisomer of allyloxy series was found to be active (paper I). Remaining stereoisomers (stereocenters are denoted by stars in Figure 8) are inactive as was determined both by computational simulations and experimental binding assays. The same stereochemistry was predicted to be active in the benzyloxy compound series.

23

Figure 9: The calculated binding free energies, for eight different plasmepsin II inhibitors, reproduce the experimental values within approximately 1 kcal/mol (dashed line).

The most important interactions between the inhibitors and Plm II were pre-dicted to be with the catalytic aspartic acids. Both hydroxyls in the ligands were donating hydrogen bonds to the charged aspartate and the P1 hydroxyl was also accepting a hydrogen bond from the protonated aspartic acid. There was also an additional interaction between the P1 hydroxyl and a water molecule, which was making a hydrogen bond network with Thr217 in the enzyme.

Having determined the favoured stereochemistry of the inhibitors, the next step was to improve the overall binding. First the interactions in the P2/P2positions of the dihydroxyethylene scaffold were optimized (Figure 10). The substitution of the valines by (1S,2R)-1-amino-2-indanol gave better binding,

Gbind= 9.1 kcal/mol compared to Gbind= 6.8 kcal/mol, respectively. Thus, the resulting compound 4 (Table 1 in paper II) gained 2.3 kcal/mol in the binding free energy. The main contribution to the higher affinity of 4, as compared to 3c, was due to an increase in the non-polar component of the free energy of binding.

24

Figure 10: Vinyl sidechains were substituted by benzyl and valine residues are exchanged for indanol in an attempt to improve binding. (Although only one valine and one vinyl are exchanged in the figure, the remaining valine/vinyl undergoes also the equivalent substitution.)

Second, in paper II different substitutions reciprocal to the scissile bond were tested, e.g,. vinyl side chains were exchanged by different aromatic systems (Figure 11). The 4-acetylphenyl extension was predicted to be the best substitution (13 in paper II) both theoretically ( Gbind= 12.2 kcal/mol) and experimentally ( Gbind= 11.3 kcal/mol). Even more interesting it proved efficient in impairing parasite growth in human erythrocytes with 78 % inhi-bition at 5 M. The other substitutions were also successful improving bind-ing to plasmepsin II relative the initial compound.

25

Figure 11: P1/P1 positions were varied in paper II by introducing different aromatic side chains to optimize binding to Plm II. (Both vinyl side chains are replaced by same type of extension.)

Comparing the calculated interaction energies for these compounds the ob-served trend was that the major contribution to the binding free energy was due to non-polar interactions, while the electrostatic component was gener-ally small and not always favourable. The aromatic substituents were longer than the initial benzyloxy and allyloxy sidechains but were nonetheless well accommodated in both the S1 and the S1 binding pocket. The P1 extension was aligned along S1-S3, thus displacing the indanol moiety towards the S2 pocket. The S1 subsite showed to be more flexible, allowing the P1 of the inhibitor to reach through it towards the solvent.

In paper II an entirely new approach was also applied to the inhibitor devel-opment by modifying the amide bond between the P1 and P2 side chains in inhibitor 6, into the methylamine compound 7. In this way the amine in the inhibitor would be protonated in the acidic food vacuole to a greater extent than outside, leading to a higher concentration of the drug at its target en-zyme while also trapping it there. Unfortunately, the inhibitor lost its activity against plasmepsin II and from MD simulations it was concluded that this was related to the poor solvation of the amine in the active site when com-pared to solution. The inhibitor was assumed to be charged in solution, but in the enzyme both the charged and uncharged forms were considered. The protonated form resulted in better binding energy than the unprotonated one, but the enzyme was not able to stabilize the charge sufficiently. In principle, the positive charge of 7 should interact with the negatively charged catalytic aspartate (Asp214 in Plm II), mimicking the protonated amine that is tran-siently present during peptide cleavage. According to our calculations, the amine of compound 7 was located 5 Å away from the Asp214, which is clearly not enough for stabilization of the charge. To our surprise when

26

tested on human erythrocytes compound 7 inhibited parasite growth by 50 % at 5 M despite having no activity in either Plm I or Plm II assays.

In conclusion, binding affinity calculations in papers I and II have made it possible to understand and improve inhibitors against Plm II. Several inhibi-tors based on the dihydroxyethylene scaffold were highly potent against both Plm I and Plm II, and also against malaria parasite infected red blood cells.

Histo-aspartic protease (paper III)Histo-aspartic protease (HAP), initially identified as plasmepsin III, attracted the attention of the research community as a potential drug target, but also as an enzyme with a novel type of active site [56, 57]. The neutral aspartic acid that is involved in the stabilization of the oxyanion of the tetrahedral inter-mediate in Plm II is mutated to a histidine residue in HAP (Plm II D34H). The proposal that the HAP peptide cleavage reaction proceeds in accordance with an aspartic protease type reaction mechanism was controversial, be-cause HAP inhibition is achieved both by the aspartic protease inhibitor pep-statin A and the serine protease inhibitor PMSF [56]. This is an interesting biological problem, and also highly challenging from the computational chemistry standpoint.

To address the question of whether HAP is an aspartic-type protease, several computational techniques were used in paper III: homology modeling, auto-mated docking and QM/MM simulations. Since there was no known struc-ture of the enzyme, a homology model was built with plasmepsin II as a template. The model was constructed by the automated modeling server, SWISS-MODEL [58], against four different Plm II structures: 1LEE, 1LF2, 1LF3 and 1LF4 [19, 36]. The high sequence identity (around 60 %) made it possible to build a homology model that is sufficiently reliable because the corresponding structures have the same fold of the core region [59]. The ob-tained model was equilibrated by molecular dynamics simulations, inside a 44 Å sphere centred at the active site, to release the strains introduced during model generation.

Having a model structure of HAP, the next step before investigating the en-zyme catalytic reaction, was to establish a substrate conformation in the ac-tive site. The substrate was a six amino acid peptide RMF LSF (where denotes the scissile bond) that was used to measure the experimental rate for the HAP catalyzed peptide cleavage. Docking studies, using AutoDock3 [60],produced two highly probable solutions. In the first one the substrate was in an extended conformation, while in the second one phenylalanine and me-thionine sidechains exchanged binding pockets. The extended peptide con-formation is generally the expected one, because most inhibitors that have

27

been crystallized in plasmepsin enzymes are also in the extended conforma-tion (cf. Figure 4 and Table 1, and references related to the peptide-based inhibitor pepstatin A). Remarkably the second conformation was very simi-lar to the corresponding one in the hemoglobin structure [61], as the peptide sequence used in enzyme assays was derived from the part of the hemoglo-bin sequence that is initially cleaved by plasmepsins [56, 62]. Since the hemo-globin-like conformation corresponds to one of the possible solutions it was found appropriate to include it in the reaction mechanism calculations.

MD/FEP/EVB molecular simulations were used to investigate the reaction mechanism of HAP. This approach is equivalent to that used for the aspartic proteases, which has been described in the previous chapter. The free energy of formation of the transient tetrahedral intermediate (TI) for the extended conformation was simulated along stepwise and concerted paths (Figure 3, and Fig. 2 in paper III). In the stepwise pathway a hydroxide ion is fully formed, while for the concerted pathway the hydrogen abstraction from wa-ter is simultaneous with the water attack on the scissile bond. The free ener-gies along the stepwise and concerted paths were also determined for the breakdown of TI. In analogy with the TI formation along the stepwise path, the scissile nitrogen is fully protonated before the peptide bond dissociates, while during the concerted reaction the process is simultaneous.

For both stepwise and concerted reaction pathways for the extended confor-mation, the TI formation was found to be rate limiting (Figure 2 in III). Mo-reover, the stepwise pathway was somewhat more favourable than the con-certed one with activation energies of 17 kcal/mol and 20 kcal/mol, respec-tively. Additional calculations were conducted to investigate the reaction mechanisms with the catalytic histidine neutral or charged. The concerted mechanism of the tetrahedral intermediate formation with the neutral his-tidine was associated with a reaction free energy barrier of approximately 26 kcal/mol. This is 6 kcal/mol higher than the corresponding reaction with the positive histidine thus making it highly unlikely. The protonation of the scis-sile nitrogen by charged histidine also showed high barriers (>25 kcal/mol), thus ruling it out as a possible reaction mechanism. The concerted TI forma-tion and breakdown were also calculated for the hemoglobin-like conforma-tion. Remarkably the reaction barrier was of the same height as for the ex-tended conformation (19 kcal/mol). However, by comparing the internal energies for the six amino acid substrates, it was found that the extended conformation was significantly favoured.

In summary, homology modelling, automated docking and MD/FEP/EVB methods have been used successfully to determine the reaction mechanism of HAP. The substrate bound in the extended conformation in HAP was predicted to be catalyzed by the aspartic protease mechanism with an activa-tion free energy corresponding to the experimental rates.

28

COLD/HEAT ADAPTATION

IntroductionTemperature adapted enzymes have to be stable at their working temperature with function still retained. Comparison of different protein homologues with different temperature optima has identified several features that are characteristic for temperature adapted enzymes. For example, an increase in the thermal stability is often accompanied with a higher number of hydrogen bonds and ion pairs. Some residues, such as prolines, are preferred in heat adapted enzymes due to their rigidity, while glycine, the residue with the most flexible backbone, appears to be included in cold adapted enzymes to counteract freezing by introducing additional flexibility. Insertion of loops also makes structures less compact with decreased thermal stability as a re-sult. [63, 64] Several other characteristics of temperature adaptation include different proportions of exposed/buried polar/non-polar surface area as de-termined from the folded structure and the extended sequence, structure compactness and the presence of cavities. In some cases proteins consist of several subunits and the interface between them can be manipulated to ob-tain different thermostabilities. [63-65]

The temperature adaptation of enzyme function affects both substrate bind-ing and catalytic steps. Substrate binding can be alleviated in cold adapted enzymes by increasing the size of the binding pocket and increasing the ac-cessibility to the active site [63]. Optimizing the electrostatic component of the substrate interactions with the enzyme can also lead to better catalytic efficiency at low temperatures [66]. However, optimization of catalysis to cold and heat, in general, is not well understood. It has been proposed that the contribution of the activation enthalpy and entropy to the activation free energy changes between normal enzymes and their cold adapted counterparts [67]. The most widespread hypothesis is that flexibility changes is the major effect behind temperature adaptation of catalysis [68].

29

Catalysis in temperature adapted citrate synthases (paper VI)Citrate synthase (CS) is part of the citric acid cycle and converts acetyl-coenzyme A (AcCoA) and oxaloacetate (OAA) to citrate. The enzyme is present in solution as a dimer in the open conformation that upon substrate binding converts to the catalytically active closed form [69]. Citrate formation consists of proton transfer, condensation and hydrolysis steps. An aspartate residue functions as a general base and abstracts a proton from the methyl group of AcCoA. The high energy enolate intermediate then attacks the car-bonyl carbon of the OAA, yielding citryl-CoA through a Claisen-type con-densation reaction. The citrate is formed as the end product after the hy-drolysis reaction. [69, 70]

Figure 12: Formation of citrate from oxaloacetate and coenzyme A by citrate syn-thase proceeds through three separate reactions: enolate formation, Claisen-type condensation and hydrolysis.

The goal of paper VI is to establish if there is any difference in the catalytic properties for the three temperature adapted citrate synthase (CS) enzymes. Free energy profiles for the keto-enol isomerization reaction step (Fig. 3 in VI) were generated by MD/FEP/EVB simulations for mesophilic S. scrofa (pig) [70], psychrophilic Arthobacter [65] and hyperthermophilic P. furiosus [71]

citrate synthase. The keto-enol isomerization reaction has been investigated previously with the same methods in glyoxalase I, triosephosphate isomerase and ketosteroid isomerase [72]. The calculations presented here are also in agreement with the mechanism of mesophilic citrate synthase as previously established with different quantum mechanical/molecular mechanics (QM/MM) methods [73-77].

The mesophilic citrate synthase (CSm) and the psychrophilic citrate synthase (CSp) have temperature optima at 55 C and 31 C, respectively [78]. The hyperthermophilic counterpart (CSh), in contrast, has an optimal working temperature that is beyond 90 C [79]. Only the proton transfer step was de-termined in paper VI because is was found to be rate limiting or very close

30

to the rate-limiting step in the pig citrate synthase [80] and is the situation is supposed to be the same for the cold adapted citrate synthase. In the hyper-thermophilic CS the proton abstraction step is probably faster than the sub-sequent reaction steps (condensation or hydrolysis), in analogy with the ther-mophilic citrate synthase from Thermoplasma acidophilum [80]. The X-ray crystallographic structures have been determined in the closed form for the CS enzymes investigated, which is essential for catalysis.

Electrostatic stabilization was found to be the most important effect for the catalytic rate enhancement. The stabilization of the transition state relative to the water reaction was on average 14 kcal/mol for the CSp, CSm and CSh

enzymes (Figure 3 in paper VI). The average electrostatic stabilization per-taining to this free energy stabilization is 24 kcal/mol (Table 2). In addition the reorganization energy reduction contributed about 20 % to the stabiliza-tion of the transition state (Figure 6 and Figure 8 in paper VI). The favour-able effect of electrostatics and reorganization energy on the activation bar-rier in enzymes is a consequence of the preorganized active site. The dipoles in the active site of an enzyme are much more firmly held than in water due to the protein 3D structure, and are able to interact more strongly with the transition state. The breaking and reforming of the hydrogen bond network during the reaction in water therefore results in less stabilization of the tran-sition state. [42, 53, 81]

Most importantly there is a significant difference in the electrostatic stabili-zation of the TS for these three enzymes. Psychrophilic CS stabilized the TS less efficiently than the mesophile CS and the same holds for the mesophile CS relative to the thermophile enzyme ( 27.9, 33.6 and 37.0, respectively in Table 2). The trend points towards psychrophilic CS being the most ‘wa-

Table 2: Average electrostatic activation (potential) energy in kcal/mol incitrate synthase.

‡elU

Energya ‡elU EVB-p EVB-w intraEVB-

water -9.2 -3.9 33.6 -38.8 CSp -27.9 16.8 -6.3 -38.5 CSm -33.6 5.5 -2.3 -36.8 CSh -37.0 9.8 -9.5 -37.2

a Average potential energy (U) components: el electrostatic; EVB-p interaction of the EVB region with solute atoms (i.e., the protein and parts of the substrates notincluded in the EVB region); EVB-w interaction of the EVB region with solvent(water); EVB-intra interaction between the atoms in the EVB region.

31

ter-like’. This would be in line with the flexibility hypothesis which states that the flexibility is correlated with the thermostability. Since the psychro-philic enzymes are least stable compared to the high-temperature homo-logues, they are also assumed to be the most flexible ones. However, calcu-lating the root mean square fluctuations (RMSF) for the six residues and three water molecules in the active site did not support the flexibility effect on catalysis, at least as far as the active site is concerned (Figure 9 in paper VI). Still, flexibility differences may be present in the other regions of en-zyme that are more important for thermostability.

In summary, MD/FEP/EVB methods have been used in paper VI to connect the transition state stabilization in three citrate synthase homologues, that are optimized to work at different temperatures, to electrostatic stabilization. Fine tuning of electrostatics may thus be one of the mechanisms that explain temperature adaptation of enzyme catalysis.

32

EPILOGUE

Plasmepsin II inhibitor design (papers I, II and V)

Plasmepsins, the hemoglobin degrading enzymes of the malaria parasite, and the dihydroxyethylene based inhibitors are presented in papers I and II. Thisinhibitor development project has considerable impact on ongoing research in the area. The plasmepsin inhibitors have thus been optimized further in several additional projects by computational (LIE calculations, ligand dock-ing and homology modelling) and experimental methods (paper V):

Amide replacement by diacylhydrazine or 1,3,4-oxadiazole was per-formed to protect inhibitors from enzyme proteolytic activity and improve the absorption, delivery, metabolization and excretion (ADME) profiles [82].

Macrocyclic compounds were made in an effort to mitigate the loss of conformational freedom upon binding to the receptor, but also to protect the amide bond from proteases [83].

Some of the above mentioned inhibitors of Ersmark et al. [82, 83] were also found to be potent against plasmepsin IV. The selectivity and activity against plasmepsin IV was explained as the result of better shape complementarity with P2 side chain with the enzyme. The corresponding subsite in Plm II was much less defined due to the motion of Met75 [84].

The future of plasmepsins as sole drug targets against the malaria parasite has recently been cast in doubt [85]. Liu and colleagues raised the question whether plasmepsins and falcipains, two predominant families in the malaria parasite food vacuole that are responsible for hemoglobin catabolism, are having overlapping functions. In conditions simulating the parasite’s blood life stage, where it is dependent on hemoglobin as the external amino acid source, the doubling time for the triple knockout falcipain-2, plasmepsin I and plasmepsin IV, was longer then for either plasmepsin I/plasmepsin IV knockout or falcipain 2 knockout. Furthermore, the aspartic protease inhibi-tor pepstatin A was highly potent against the falcipain-2 knockout. Thus the

33

authors concluded that plasmepsin inhibition is a viable alternative only when falcipain inhibition is considered concomitantly.

Nevertheless, plasmepsin II directed inhibitor development, as described in papers I, II and V, still appears to be promising in view of the fact that com-pound 13 in paper II was potent against parasite infected red blood cells with 78 % growth inhibition. Similarly, the parasite growth inhibition of 50 % in read blood cells was achieved by compound 7 (in paper II) but its inhibition mechanism remains to be explained. This compound did not have any effect on plasmepsin II inhibition and probably affects another essential enzymatic pathway.

Papers I, II and V provide a good example of how computational and ex-perimental methods can be combined to develop inhibitors against drug tar-gets and understand the interactions behind ligand binding.

Histo-aspartic protease

In paper III free energy profiles were calculated for peptide cleavage by histo-aspartic protease (HAP). The HAP research was motivated by the great attention that plasmepsins were receiving as drug targets and by the lack of structural information for HAP. The main reason that there was no HAP structure was, and still is, due to the lack of functioning recombinant expres-sion of the HAP protein. However, in a recent paper Xiao and co-workers have hopefully solved the expression problem and a better characterization may be underway [86].

A combination of computational techniques in paper III, including homol-ogy modelling, automated docking, and MD/FEP/EVB simulations was nec-essary to explore the reaction mechanism. With these methods the substrate binding mode was predicted as well as the free energy profiles for the en-zyme reaction. A significantly higher level of theory was thus applied com-pared to the previously published works of Berry et al. and Andreeva et al.[57, 87]. These authors made use of relatively simple methods, homology mod-elling and standard molecular dynamics simulations, respectively, to gener-ate HAP 3D models. Based on their 3D homology model Berry et al. pro-posed that HAP was a novel type of protease with a histidine and an aspar-tate forming the catalytic dyad in agreement with our conclusion. On the contrary, Andreeva et al. [87] proposed that a serine protease type of mecha-nism in spite of aspartic protease fold was the more probable of the two pos-sible alternative mechanisms initially proposed by Banerjee et al. [56], i.e.aspartic protease versus serine protease. This hypothesis was, however, only based on apoenzyme simulations and by visual inspection of the resulting active site.

34

In summary, paper III strongly supports the aspartic protease type mecha-nism which currently also is the favoured mechanism. The combination of computational methods used to determine the reaction mechanism in paper III is possibly the first of its kind and may find further use in the future.

Enzyme catalysis

Different factors have been put forward to explain the catalytic power of enzymes and some of the most popular theories include:

Electrostatics of the preorganized active site [81]:the protein fold provides the active site with dipoles that are orien-tated to interact specifically with the transition state.

Dynamical effects [88]:enzyme specific vibrational modes have evolved to push the sub-strate into the transition state.

Low-barrier hydrogen bonds (LBHBs) [89]:partially covalent hydrogen bond character in the enzymes is respon-sible for catalysis.

Near-attack conformations (NACs) [90]:reactants are positioned closer to the transition state in enzyme than in the water reaction.

In papers IV and VI the origin of the catalytic effect was found to reside in electrostatic stabilization and the preorganized environment of the active site. In addition the dynamical effect was examined as Piana et al. and Cas-cella et al. have implied that the catalytic effect of HIVP may depend on the enzyme motion [91, 92].

The simulation of the tetrahedral intermediate formation in paper IV with enzyme motions restrained to different points along the previously unre-strained reaction path did not have any considerable effect on the height of the activation barrier in the HIVP/KILFLD system. This is natural, since enzyme dipoles are still able to follow and stabilize the reaction, while fluc-tuations are significantly lower. On the contrary, the two simulations where the enzyme structure was restrained to either reactant or tetrahedral interme-diate states during the reaction resulted in an increase and decrease of the activation barrier in comparison to the unrestrained reaction, respectively. The increase in the activation barrier is similar to that found by Piana et al.[92] and can be explained by the fact that the active site residues are interact-

35

ing favourably with the substrate, but are not positioned to adequately stabi-lize the transition state.

The computational approach of paper IV gives a comprehensive description of the transition state stabilization by aspartic proteases, and connects this stabilization to electrostatics and the reduction of the reorganization energy compared to uncatalyzed peptide bond cleavage in water.

Cold/heat adaptation of catalysis

In paper VI the effect of temperature adaptation on keto-enol isomerization step was investigated for three different citrate synthase homologues. Elec-trostatic stabilization of the transition state showed a connection with the temperature adaptation. In contrast, no support was found for the flexibility hypothesis that has been put forward as a major contribution to the catalytic efficiency in cold adapted enzymes [68, 93]. However, the change of flexibility in other regions of the enzyme, outside the active site, may still be important for thermostability. While in paper VI only three citrate synthase homo-logues are investigated the results may suggest a general temperature adapta-tion mechanism. Of course this has to be verified in other systems before we can come to any definite conclusion. It would also be possible to carry out calculations on citrate synthase from Thermoplasma acidophilum and Sul-folobus solfataricus to verify the hypothesis. These citrate synthase enzymes have temperature optima between those of mesophilic and hyperthermo-philic citrate synthase. Unfortunately, the crystal structures of T. acidophi-lum and S. solfataricus citrate synthase are only available in open configura-tion [94, 95] and additional modelling is required before such calculations can be undertaken.

The investigation in paper VI of the detailed energetics behind catalysis points thus clearly towards electrostatics as the main effect of temperature adaptation on the transition state stabilization in citrate synthases.

Conclusion

Molecular simulations of the peptide cleavage reaction in plasmepsin II, histo-aspartic protease, cathepsin D and HIV protease, the keto-enol isom-erization reaction in citrate synthase, and of plasmepsin II inhibitor binding presented in this thesis are a contribution to our understanding of enzyme catalysis and inhibition. New advances in theoretical methods as well as development of the computer hardware and software hold considerable promise for the future of these methodologies.

36

THEORETICAL METHODS

The following chapter contains a description of the theoretical and computational methods used in this work.

Force fields An empirically derived potential energy function that describes interactions present between the atoms, in a molecule or separate molecules, is usually called a force field (FF). Popular force fields for the simulation of macromolecules are AMBER [96] (used in paper VI), CHARMM [97, 98], OPLS-AA [99] (used in papers III, IV, and V) and GROMOS [100, 101] (GROMOS87 [100] was used in papers I and II).

The non-bonded interactions in a FF [96-101] are a sum of electrostatic and van der Waals contributions

12 60

14

i j i j i jnb

atom atomij ij ijpairs pairs

q q A A B BU

r r r (4)

where q is the partial charge situated on the atom centers i and j, rij is the distance separating them, 0 denotes the electric permittivity of vacuum, and A and B are the Lennard-Jones parameters that depend on the chemical nature of the interacting atoms.

The bonded interactions (bn) are the sum of bond, angle and torsion terms:

20

20

( )

( ) [1 cos( )]

bn b ijbonds

ijk ijklangles torsions

U k r r

k k n (5)

kb, k , k are the bond-stretching, angle-bending and torsional force constants, re-spectively. r0 is the equilibrium bond length and 0 the equilibrium angle. is the angle between atoms i, j and k. n is the number of minima per full turn of the torsion angle , and is the location of the first barrier. The torsion is defined for atoms i j k l as an angle between planes i j k and j k l (Figure 13).

37

Figure 13: A torsion is defined between atoms i j k l and an improper tor-sion as j k l m (i, j, k, l, and m refer to heavy atoms in green).

The potential energy model for torsion angles is sufficient for describing most cases, but sometimes atoms that are, for example, in a plane are bent out of it. Thus to hold atoms in correct orientation an improper torsion angle between planes j k l and k l m (i.e. improper) is defined.

SolventThe simulations were carried out by immersing molecules of interest in a water sphere. The representation of solvent is thus as important as the description of the macromolecule for the analysis of chemical properties of biological systems. The SPC [102] and TIP3P [103] water models were employed in the current thesis. Water at the boundary of the simulation sphere is subjected to radial and polarization re-straints [104, 105]. In cases when water sphere is smaller than the macromolecule, all atoms outside the sphere are restrained with the positional restraints of 100 kcal mol-

1 Å 2.

Docking

AutoDock3 [60], Dock [106], FlexX [107], Glide [108, 109] and GOLD [110] are examples of docking tools that are used routinely for the prediction of bound ligand conforma-tions. In paper III AutoDock3 was used for binding mode prediction of the six amino acid substrate. It determines the conformation of a ligand by docking into pre-calculated grid maps. The grid maps describe van der Waals, hydrogen bonding, electrostatic and solvation potentials at each point of the grid. From the grid maps, the different conformations are assessed with the scoring function,

38

12 6 12 10, ,

( ) ( ) ( )ij ij ij ijvdw hb

i j i jij ij ij ij

A B C DG G G E t

r r r r

2

22

. ,( )

( )

ijri j

el tor tor solv i j j ii j i jij ij

q qG G N G S V S V e

e r r (6)

where empirically derived G factors are used as weights. These have been param-eterized on a training set of 30 protein ligand complexes [60]. The first and third terms are the electrostatic and van der Waals contributions as described previously (cf. eq. 4). The hydrogen bonding term (hb) is weighted with a function E depending on the angle t between the donor, hydrogen and acceptor atoms. The above equation also includes a torsion-dependent term (tor) that is proportional to the number of sp3

hybridized atoms. This term is used for describing the unfavourable restriction of motion upon transfer from water to the active site of a receptor. The last term is the desolvation estimated upon transfer from water to the protein binding pocket. It is estimated as a sum of solvation parameters and approximated atomic volumes scaled with an exponential function. Conformational space is searched using a genetic algorithm combined with a local search method.

Docking techniques are often extremely time-consuming and less reliable for the docking of large ligands. Thus in papers I, II, IV and VI the inhibitors in the X-ray crystallographic structures were used as templates for the initial, manual placement of the corresponding atoms in the substrate/inhibitor. Manual superposition is natu-rally more biased then the automated docking, but because the complexes were used as the starting structures for MD simulations, their binding mode is further opti-mized during equilibration.

ScoringScoring functions were used as an alternative method to predict inhibitor binding affinities. Scoring methods are extremely fast, but at the disadvantage of accuracy and are often used for the relative ranking of inhibitors. The Chemscore scoring function [111] was initially used to select the most potent stereoisomer in paper II. It consists of five different terms that account for hydrogen bonding, metal ions, lipo-philic interactions and the number of rotatable bonds. Each function consists of several empirical constants ( Gi) that scale linearly with each type of interaction:

bind H bond H bondG G N

0metal metal lipo lipo rot rotG N G N G N G (7)

Other scoring functions that have been widely used are the Drugscore function by Klebe [112], X-score function by Wang [113] and Goldscore [110, 114]. The X-score func-

39

tion was used by Gutiérrez-de-Terán et al. for the evaluation of plasmepsin IV in-hibitor binding [84] and is discussed in the review paper V. The X-score function has the form [113]

0bind vdW H bond rot hydrophobicG G G G G G (8)

where GvdW determines the van der Waals energy between ligand and protein, GH-bond is a directional hydrogen-bond term, Ghydrophobic represents hydrophobic

contributions and Grot describes the rotational entropy penalty for ligand transfer from solution to protein binding site. G0 is as in eq. 7 a regression constant.

Molecular Dynamics Docking and scoring techniques represent a rather simplified view where free ener-gies are assigned to single snapshots rather than thermal averages of the system. Molecular dynamics (MD) [115] was thus used in this thesis to propagate the investi-gated systems through time with respect to the forces that act on individual atoms. If we consider N atoms, with spatial positions r1(t), ..., rN(t), acting under some kind of a potential energy U(r1, ..., rN), it is possible to determine the forces F1, ..., FN influ-encing their motion:

1( ( )Ni

i

U r rFr

(9)

Newton’s law of motion is integrated during the MD simulations with successive configurations of the system as a result. A number of algorithms exist that deal with finite time steps and truncates the series expansion at different terms. The algorithm used for the MD simulations in this thesis is the leap-frog version of the Verlet algo-rithm as implemented in the simulation package Q [116]. Here it is assumed that the velocity, constant over a finite time step, is calculated at the midpoint of the simula-tion

( ) ( ) ( )2 2i i it tt t t tr r r (10)

with new positions at next time step t given by:

( ) ( ) ( )2i i itt t t t tr r r (11)

The acceleration in eq. 10 is obtained from Newton’s second law. The position equa-tion can be integrated with the velocity equation into:

40

2( ) ( ) ( )2

ii i i

i

tt t t t t tmFr r r (12)

The leap-frog version of the Verlet algorithm has the advantage of allowing easy temperature control since the velocity explicitly enters into the algorithm [115]. The initial velocities at time t equal to zero are determined by random selection from the Maxwell-Boltzmann probability distribution, p, at the simulation temperature

2

( ) exp2 2

i i ii

B B

m mpk T k T

rr (13)

where T is the temperature and kB is the Boltzmann constant.

Statistical mechanics Statistical mechanics [115, 117] connects MD simulations with experimentally deter-mined properties. Averaging over a large number of systems that are in different microscopic configurations gives the macroscopic property formulated in terms of an ensemble average:

( , ) ( , ) N NA A d dp r p r p r (14)

The desired property A is a function of momenta pi(t), ..., pN (t) and positions ri(t), ..., rN(t) of N particles at time t. The angle brackets < > indicate an ensemble average (mean value) and is the probability density, i.e. the probability of finding a con-figuration with the momenta p and positions r. There are different statistical ensem-bles depending on the thermodynamic properties that are held constant. In the ca-nonical ensemble where number of particles N, volume V and temperature T are constant, the probability density takes the form of the Boltzmann distribution

3

exp( ( , ) /( ))( , )exp( ( , ) /( )

BN N N

B

E k Th E k T d d

p rp rp r p r

(15)

where the denominator is the canonical partition function Q. The energy, E is the sum of kinetic and potential energies, which are independent and thus separable. The non-kinetic part can be expressed as

exp( ( ) /( )) NBZ U k T dr r (16)

where U is the potential energy and Z is the configurational integral. After integrat-ing a property A over all possible conformations the ensemble average is deter-

41

mined. The fundamental principle in statistical mechanics, the ergodic hypothesis, states that the ensemble average is equal to the time average for a single system:

0

1lim ( ( ), ( )) ( , ) ( , ) N N

tA t t dt A d dp r p r p r p r (17)

This means that given an infinite amount of time the system will cover all the possi-ble conformations and averaging over a trajectory of the system is equivalent to the averaging over the ensemble. In MD the time average over an observable A is ob-tained by calculating the statistical average of a property sampled over a trajectory, where M is the number of time steps and A(p(ti),r(ti)) is the instantaneous value of Aat the time step ti. If M goes to infinity the ergodic hypothesis is equivalent to the statement

1

1lim ( ( ) ( ))M

i iM iA A t t

Mp r (18)

i.e. a single numerical trajectory can be used for calculating the MD ensemble aver-age of an observable.

Free energy perturbation (FEP)In thermodynamics, the free energy is measured in two ways depending on what restraints are imposed on the system. If the pressure and volume are held constant Gibbs free energy and Helmholtz free energy are determined, respectively. During ligand binding experiments both volume and pressure are approximately constant for most macroscopic systems and the free energy for this process is referred in the rest of the thesis as F. A way of obtaining free energy from a simple chemical reaction of ligand, L, binding to a receptor, R, goes by the dissociation constant Kd

0 ln /bind dF RT K c (19)

where F0bind is the standard free energy of binding and c is the standard state con-

centration. The property F that is calculated from MD simulations, in this case the free energy, is derived from statistical mechanics by

lnBF k T Q (20)

where Q is the canonical partition function [117]. The difference in free energy, F,between two different states X and Y, after substitution of the canonical partition function with the configurational integral Z, is defined by

42

1exp( )

lnexp( )

NY

NX

U dF

U d

r

r (21)

where U(r1, ..., rN) is the potential energy and =1/ kBT. To get the expression in a form that can be applied in molecular simulations it has to be transformed to a useful form with a single ensemble average (which is done by substituting exp( VX ) · exp(+ VX ) = 1 in the numerator):

1 ln exp( )X

F U (22)

U is the difference of the potential energies for the two different states X and Y and the subscript X denotes averaging over the configurations representative of the initial state. The previous equation is called the free energy perturbation formula and is attributed to Zwanzig [118].

The generation of free energy profiles is accomplished by combining the free energy perturbation (FEP) method with the molecular dynamics (MD) simulations. FEP is used to drive the system between two different states. Often the energy difference between these end states is too large for efficient sampling, and transformation is conducted by using several intermediate mapping potentials:

1 2m m m

Y XU U U and 1 2 1m m (23)

The mapping vector m, defines a linear combination between the potentials, chang-ing between the values (1,0) for reactants and (0,1) for products, with the constraint that the sum of 1 and 2 is always equal to 1. The free energy is obtained by modi-fying the FEP equation [118] (eq. 22) to include n intermediate steps

11

0 10

( ) ( ) ln exp ( )n

n n m m mm

F F U U (24)

where =1/kT and the average, denoted by broken parentheses, is evaluated on the mapping potential surface Um

[119].

Linear interaction energy (LIE) method The free energy perturbation (FEP) method presented in the previous section is usu-ally successful in inhibitor binding predictions only when the difference between the two potentials of interest is small. In comparison to FEP, scoring functions are usu-ally not sufficiently accurate for detailed investigation of ligand binding and lead optimization. Alternatively, several semi-empirical methods have been proposed for the determination of absolute binding free energies, for example MM-PBSA (mo-lecular mechanics/Poisson-Boltzmann/surface area) [120], LRA (linear response ap-

43

proximation) [121, 122] and LIE [123]. In papers I and II the LIE method was used for binding affinity prediction of several novel Plm II inhibitors. The binding free en-ergy is determined by performing two separate simulations: i) the interaction energy of a solvated ligand is calculated in the first simulation, ii) in the second calculation the ligand surroundings interactions are determined from the solvated pro-tein ligand complex.

The ligand binding free energy is estimated as a sum of the difference of the polar and non-polar interactions from the two simulations. These differences are scaled with the appropriate constants that are determined empirically ( ) or theoretically ( LIE)

bindF

vdw vdw el ell s l s l s l sbound free bound free

U U U U (25)

where el and vdW are electrostatic and van der Waals interactions of the ligand with surroundings (l s), respectively, in the free state (in water) and bound to protein.

The theoretically derived scaling constant LIE that connects the electrostatic interac-tion with the free energies is derived by expansion of the FEP equation (eq. 22). Theoretically this value equals 0.5 when the electrostatic linear response assumption holds. A revised model divides inhibitors in four classes: charged, dipolar with no hydroxyl groups, dipolar with one hydroxyl group and dipolar with two or more hydroxyl groups and assigning them values of LIE = 0.5, LIE = 0.43, LIE = 0.37 and

LIE = 0.33 respectively [124]. Since the dihydroxyethylene inhibitors in papers I and II have at least two hydroxyl groups the scaling constant of 0.33 is applied except for the charged compound with 0.5 scaling. The van der Waals scaling constant is, in contrast to LIE, derived empirically. It has been found that the solvation free energy scales linearly with molecular size measures, which also motivated the use of the van der Waals energies for the calculation of the non-polar binding free energy [123]. The optimization of on the set of 18 complexes gave a value of 0.18 [124],which has been used in papers I and II.

Empirical valence bond (EVB) methodThe empirical valence bond (EVB) method, used in papers III, IV and VI, has a background in the quantum mechanics (QM) valence bond theory [50, 125]. The ground state potential energy surface, Eg, for a two state model (e.g. reactants and products) is obtained as non-zero eigenvalue solutions of the secular equation:

2 21 2 1 2 12

1 12 2( ) ( ) 4gE H (26)

44

The connection with MD simulations goes by force field like potentials i that de-scribe the end points of the reaction. They are the sum of the bonded interactions present within (or across) the reacting region and the nonbonded nb interactions

( ) ( ) ( ) ( ) ( ) ( ), ,

i i i i i iii i bond angle torsion nb rr nb rs ssH U U U U U U (27)

where rr is the interactions within the reacting region, rs is the interaction of the reacting region with the surroundings and ss is the interaction of the surrounding water/enzyme complex.

A key feature of the EVB method is the presence of the gas phase energy of the reacting fragments at the infinite separation, (i), and the coupling constant, Hij,describing the off-diagonal matrix elements. These parameters are used to empiri-cally fit the simulated water reaction to the experimental or ab initio data for a suit-able reference reaction in solution. The off-diagonal coupling constant is described in papers III, IV and VI by constant value [126]. As an alternative, different expo-nential forms have also been used successfully [50, 125, 126]. When the water reaction is parameterized, the same reaction is simulated in the enzyme active site without changing the fitted parameters. Through this procedure errors associated with the force fields can be reduced and the effect of the enzyme environment is easily ob-tained. Finally, the umbrella sampling expression makes it possible to calculate the free energy profiles F(X) corresponding to trajectories moving on the actual ground-state potential, Eg(X), (X is the reaction coordinate described by the differ-ence in the diabatic energy gap of reactants and products [50, 125, 126]).

45

SUMMARY IN SWEDISH

Forskningsresultat presenterade i den här avhandlingen inriktar sig mot lä-kemedelsutveckling inom malaria och temperaturadaptering av katalys i citratsyntasenzymet från olika organismer.

MalariaMalaria är mest utbredd i tropiska områden och man räknar med att flera hundra miljoner människor är infekterade med väldigt höga dödssiffror (flera miljoner per år). Sjukdomen sprids genom myggbett och flera projekt har inletts för att dränera stora vattenområden i hopp om att minska myggans naturliga habitat. En annan del av forskningen fokuserar på att utveckla ett fungerande vaccin. Men där har man stött på problem, eftersom malariapara-siten är en komplicerad organism och det är svårt att välja något lämpligt mål att rikta vaccinet mot. Malaria parasitens resistansutveckling mot befint-liga läkemedel är också ett allvarligt problem.

Inhibitorer som har studerats riktar sig mot malariaparasiten när den befinner sig i röda blodkroppar hos människan. Eftersom den livnär sig på hemoglo-bin, försöker man förhindra sjukdomsutvecklingen genom att blockera de enzymer som är inblandade i hemoglobinnedbrytningen. Det finns olika strategier för att designa inhibitorer som binder till ett specifikt protein. Man kan till exempel genomsöka stora databaser av kemiska föreningar och be-stämma vilken av dessa föreningar som bäst växelverkar med den aktuella receptorn. Den här typen av försök lämpar sig bäst när man egentligen inte vet särskilt mycket om själva receptorn. Alternativt kan man studera reaktio-ner som enzymer är involverade i och bestämma vilka interaktioner som är fördelaktiga under katalysen. Dessa interaktioner behåller man sedan i inhi-bitorn med ytterligare modifikationer för att göra den mer verksam.

Malariaparasitens enzym som bryter ner hemoglobin har därför undersökts med hjälp av teoretiska metoder och publicerats i artikel III och artikel IV.

46

Eftersom flera olika enzymer är involverade i nedbrytningen, har reaktions-mekanismen bestämts för två av dem: plasmepsin II (Plm II) och histo-aspartatproteas (HAP). Plasmepsin II har attraherat forskarvärldens intresse under nästan ett decennium eftersom den tillsammans med plasmepsin I är involverad i hemoglobinnedbrytningen. Forskningen har även lett till upp-täckten av plasmepsin III (HAP) och plasmepsin IV som också är involvera-de i samma process Plm I och Plm II. Den aktuella arbetshypotesen var att inhibering av dessa hemoglobinnedbrytande enzymer skulle leda till parasi-tens död.

Plm II och HAP är aspartylproteaser med 60 % sekvensidentitet med den avgörande skillnaden i deras katalytiska mekanism. Aspartylproteaser klyver peptider genom hydrolys. Vattenmolekylen aktiveras av den negativt ladda-de katalytiska aspartaten och attackerar kolatomen i peptidbindningen. Den nu neutrala aspartaten ger protonen till kvävet i peptidbindningen vilket le-der till peptiddissociering. Den största skillnaden mellan Plm II och HAP ligger i hur de stabiliserar den tetrahedralaintermediatet under reaktionen. Plm II har ytterliggare en aspartat som är essentiell för stabilisering av reak-tionen, medan i HAP uppnås stabilisering av en positivt laddad histidin. Ef-tersom HAP hade ett hittills okänt sätt att bryta ner hemoglobin på, var det intressant även inom andra forskningsområden att veta dess reaktionsmeka-nism.

Artikel I och artikel II i avhandlingen handlar om att utveckla och optimera potentiella inhibitorer mot Plm II. Dessa inhibitorer liknar korta peptider, oftast fyra till sex aminosyror långa, med små kemiska förändringar. Det är dessa förändringar som optimeras för att ge en bättre bindning. De viktigaste interaktionerna mellan de undersökta inhibitorerna och Plm II är centrerade kring de centrala hydroxylerna i inhibitorn (Figur 1).

Figur 1: Plasmepsin II inhibitorer har två centrala hydroxyler, markerade med en cirkel och utgör grunden för interaktioner med de katalytiska asparta-terna i enzymet.

47

Genom att byta ut vinyl-sidokedjorna mot olika aromatiska föreningar och valinerna mot indanol-grupper i Figur 1 har flera potenta inhibitorer genere-rats. En av dessa, som är presenterad i Figur 2, band till Plm II vid nM kon-centration och bromsade dessutom malariaparasitens tillväxt med 78 % vid test på röda blodkroppar in vitro.

Figur 2: Inhibitorn med acetylphenyl-grupperna (inom markerat område) var mest aktiv mot plasmepsin II.

Temperatur-optimering av citratsyntas Katalysen hos tre homologa citratsyntasenzymer, som är evolverade för att generera citrat vid olika temperaturer, har bestämts i artikel VI. Citratsyntas katalyserar oxalacetat och acetyl-koenzym A till citrat. Metylgruppen i ace-tyl-koenzym A deprotoneras vilket leder till att enolat bildas (Figur 3). I nästa steg attackerar enolatet oxalacetat och bildar citryl-koenzym A. Denna

48

reaktion, även kallad kondensation, är det näst sista steget innan hydrolys reaktionen som leder till citrat.

Figur 3: Katalytiskt aspartat i citratsyntas deprotonerar acetyl-koenzym A och bildar enolat.

För att studera temperatur-adaptering av enzymkatalys, har reaktionshastig-heter bestämts i citratsyntasenzymer som fungerar vid kyla, rumstemperatur och värme. Endast protonöverförningssteget har beaktats i artikel VI. Under reaktionen visade det sig att elektrostatiska interaktioner var det viktigaste bidraget till katalys jämfört med vattenreaktionen. Ännu intressantare var att elektrostatiken hade olika stabilisering i citratsyntashomologer. Bäst stabili-sering uppnåddes i den varma homologen, därefeter i den rumstempererade och sist den kalla. Hypotesen som har framförs i artikel VI är att elektrosta-tiken eventuellt kan ha samma effekt på temperatur-adaptering i andra en-zymer.

49

ACKNOWLEDGMENTS

This thesis and the research within would not have been possible without help and support from:

SupervisorProf. Johan Åqvist for trying to make a scientist out of me. Thank you for the scientific freedom and letting me find my own way.

Group members and former group members Martin Almlöf, unreal computer expert Martin Andér, illustrator and photoshop master Jens Carlsson, for always having a good opinion Martin Nervall, a perfect roommate Lars Persson, fresh blood in the group Stefan Trobro, of great imagination Fredrik Österberg, a lot of fun Isabella Feierberg, degree project supervisorBjørn O. Brandsdal, organizing the best FEBS course

Hugo Gutiérrez-de-Terán, for scientific enthusiasm Viktor Luzhkov, for chemical intuition

Everybody at xray and Beer club organizers Making the everyday life more fun.

Anna Maria Lundin (AML) and Liljewalchs foundations Providing numerous travel grants.

Micke

lim # ( ) discussionst PhD

climbing t

Family in Sweden, Germany and BiH Always being there for me.

Ana & Stanoje – najbolji roditelji!

50

K(W)B Making me see what is important.

51

REFERENCES

[1] J. K. Baird, N. Engl. J. Med. 2005, 352, 1565. [2] B. Greenwood, T. Mutabingwa, Nature 2002, 415, 670. [3] B. Greenwood, Acta Trop. 2005, 95, 298. [4] B. M. Greenwood, K. Bojang, C. J. M. Whitty, G. A. T. Targett,

Lancet 2005, 365, 1487. [5] L. H. Miller, D. I. Baruch, K. Marsh, O. K. Doumbo, Nature 2002,

415, 673. [6] R. W. Snow, C. A. Guerra, A. M. Noor, H. Y. Myint, S. I. Hay, Na-

ture 2005, 434, 214. [7] G. A. Targett, Trends Parasitol. 2005, 21, 499. [8] E. A. Winzeler, Nat. Rev. Microbiol. 2006, 4, 145. [9] D. F. Wirth, Nature 2002, 419, 495. [10] P. K. Chiang, J. M. Bujnicki, X. Z. Su, D. E. Lanar, Curr. Mol. Med.

2006, 6, 309. [11] U. Eckstein-Ludwig, R. J. Webb, I. D. A. van Goethem, J. M. East,

A. G. Lee, M. Kimura, P. M. O'Neill, P. G. Bray, S. A. Ward, S. Krishna, Nature 2003, 424, 957.

[12] J. Wiesner, R. Ortmann, H. Jomaa, M. Schlitzer, Angew. Chem. Int. Ed. 2003, 42, 5274.

[13] W. Sirawaraporn, T. Sathitkul, R. Sirawaraporn, Y. Yuthavong, D. V. Santi, Proc. Natl. Acad. Sci. U. S. A. 1997, 94, 1124.

[14] K. Chaudhary, D. S. Roos, Nat. Biotechnol. 2005, 23, 1089. [15] M. J. Gardner, N. Hall, E. Fung, O. White, M. Berriman, R. W.

Hyman, J. M. Carlton, A. Pain, K. E. Nelson, S. Bowman, I. T. Paulsen, K. James, J. A. Eisen, K. Rutherford, S. L. Salzberg, A. Craig, S. Kyes, M.-S. Chan, V. Nene, S. J. Shallom, B. Suh, J. Peter-son, S. Angiuoli, M. Pertea, J. Allen, J. Selengut, D. Haft, M. W. Mather, A. B. Vaidya, D. M. A. Martin, A. H. Fairlamb, M. J. Fraunholz, D. S. Roos, S. A. Ralph, G. I. McFadden, L. M. Cum-mings, G. M. Subramanian, C. Mungall, J. C. Venter, D. J. Carucci, S. L. Hoffman, C. Newbold, R. W. Davis, C. M. Fraser, B. Barrell, Nature 2002, 419, 498.

52

[16] T. W. A. Kooij, C. J. Janse, A. P. Waters, Nat. Rev. Microbiol. 2006,4, 344.

[17] K. Ersmark, B. Samuelsson, A. Hallberg, Med. Res. Rev. 2006, 26,626.

[18] P. J. Rosenthal, Int. J. Parasitol. 2004, 34, 1489. [19] O. A. Asojo, S. V. Gulnik, E. Afonina, B. Yu, J. A. Ellman, T. S.

Haque, A. M. Silva, J. Mol. Biol. 2003, 327, 173. [20] T. L. Blundell, J. Cooper, S. I. Foundling, D. M. Jones, B. Atrash,

M. Szelke, Biochemistry 1987, 26, 5585. [21] D. R. Davies, Annu. Rev. Biophys. Biophys. Chem. 1990, 19, 189. [22] B. M. Dunn, Chem. Rev. 2002, 102, 4431. [23] A. Fersht, Structure and mechanism in protein science, W. H. Free-

man and Company, New York, 1999.[24] L. J. Hyland, T. A. Tomaszek, T. D. Meek, Biochemistry 1991, 30,

8454. [25] L. J. Hyland, T. A. Tomaszek, G. D. Roberts, S. A. Carr, V. W. Ma-

gaard, H. L. Bryan, S. A. Fakhoury, M. L. Moore, M. D. Minnich, J. S. Culp, R. L. Desjarlais, T. D. Meek, Biochemistry 1991, 30, 8441.

[26] L. Pearl, T. Blundell, FEBS Lett. 1984, 174, 96. [27] R. Smith, I. M. Brereton, R. Y. Chai, S. B. H. Kent, Nat. Struct.

Biol. 1996, 3, 946. [28] K. Suguna, E. A. Padlan, C. W. Smith, W. D. Carlson, D. R. Davies,

Proc. Natl. Acad. Sci. U. S. A. 1987, 84, 7009. [29] B. Veerapandian, J. B. Cooper, A. Sali, T. L. Blundell, R. L. Rosati,

B. W. Dominy, D. B. Damon, D. J. Hoover, Protein Sci. 1992, 1,322.

[30] T. D. Meek, E. J. Rodriguez, T. S. Angeles, in Retroviral Proteases, Vol. 241, 1994, pp. 127.

[31] E. J. Rodriguez, T. S. Angeles, T. D. Meek, Biochemistry 1993, 32,12380.

[32] A. Bagno, G. Lovato, G. Scorrano, J. Chem. Soc., Perkin Trans. 2 1993, 1091.

[33] R. S. Brown, A. J. Bennet, H. Slebocka-Tilk, Acc. Chem. Res. 1992,25, 481.

[34] J. P. Guthrie, J. Am. Chem. Soc. 1974, 96, 3608. [35] D. B. Northrop, Acc. Chem. Res. 2001, 34, 790. [36] O. A. Asojo, E. Afonina, S. V. Gulnik, B. Yu, J. W. Erickson, R.

Randad, D. Medjahed, A. M. Silva, Acta Crystallogr., Sect D: Biol. Crystallogr. 2002, D 58, 2001.

[37] N. K. Bernstein, M. M. Cherney, H. Loetscher, R. G. Ridley, M. N. G. James, Nat. Struct. Biol. 1999, 6, 32.

[38] L. Prade, A. F. Jones, C. Boss, S. R. Bildstein, S. Meyer, C. Binkert, D. Bur, J. Biol. Chem. 2005, 280, 23837.

[39] A. M. Silva, A. Y. Lee, S. V. Gulnik, P. Majer, J. Collins, T. N. Bhat, P. J. Collins, R. E. Cachau, K. E. Luker, I. Y. Gluzman, S. E. Francis, A. Oksman, D. E. Goldberg, J. W. Erickson, Proc. Natl. Acad. Sci. U. S. A. 1996, 93, 10034.

53

[40] PDB, To be published.[41] H. Eyring, A. E. Stearn, Chem. Rev. 1939, 24, 253. [42] J. Villa, A. Warshel, J. Phys. Chem. B 2001, 105, 7887. [43] M. Garcia-Viloca, J. Gao, M. Karplus, D. G. Truhlar, Science 2004,

303, 186. [44] A. Brik, C. H. Wong, Org. Biomol. Chem. 2003, 1, 5. [45] N. S. Andreeva, L. D. Rumsh, Protein Sci. 2001, 10, 2439. [46] J. Åqvist, in Computational approaches to biochemical reactivity

(Eds.: G. Náray-Szabó, A. Warshel), Kluwer Academic Publishers, 1997, pp. 341.

[47] A. R. Fersht, R. J. Leatherbarrow, T. N. C. Wells, Biochemistry1987, 26, 6030.

[48] D. N. Silverman, C. K. Tu, X. Chen, S. M. Tanhauser, A. J. Kresge, P. J. Laipis, Biochemistry 1993, 32, 10757.

[49] M. D. Toney, J. F. Kirsch, Science 1989, 243, 1485. [50] A. Warshel, Computer modeling of chemical reactions in enzymes

and solutions, John Wiley & Sons, New York, 1991.[51] A. Warshel, T. Schweins, M. Fothergill, J. Am. Chem. Soc. 1994,

116, 8437. [52] W. J. Albery, Annu. Rev. Phys. Chem. 1980, 31, 227. [53] A. Warshel, P. K. Sharma, M. Kato, Y. Xiang, H. B. Liu, M. H. M.

Olsson, Chem. Rev. 2006, 106, 3210. [54] V. L. Schramm, Curr. Opin. Struct. Biol. 2005, 15, 604. [55] S. Marti, M. Roca, J. Andres, V. Moliner, E. Silla, I. Tunon, J. Ber-

tran, Chem. Soc. Rev. 2004, 33, 98. [56] R. Banerjee, J. Liu, W. Beatty, L. Pelosof, M. Klemba, D. E. Gold-

berg, Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 990. [57] C. Berry, M. J. Humphreys, P. Matharu, R. Granger, P. Horrocks, R.

P. Moon, U. Certa, R. G. Ridley, D. Bur, J. Kay, FEBS Lett. 1999,447, 149.

[58] T. Schwede, J. Kopp, N. Guex, M. C. Peitsch, Nucleic Acids Res. 2003, 31, 3381.

[59] C. Chothia, A. M. Lesk, EMBO J. 1986, 5, 823. [60] G. M. Morris, D. S. Goodsell, R. S. Halliday, R. Huey, W. E. Hart,

R. K. Belew, A. J. Olson, J. Comput. Chem. 1998, 19, 1639. [61] B. Luisi, N. Shibayama, J. Mol. Biol. 1989, 206, 723. [62] K. E. Luker, S. E. Francis, I. Y. Gluzman, D. E. Goldberg, Mol.

Biochem. Parasitol. 1996, 79, 71. [63] G. Feller, C. Gerday, Nat. Rev. Microbiol. 2003, 1, 200. [64] A. O. Smalås, H. K. S. Leiros, V. Os, N. P. Willassen, Biotechnol.

Annu. Rev. 2000, 6, 1. [65] R. J. M. Russell, U. Gerike, M. J. Danson, D. W. Hough, G. L. Tay-

lor, Structure 1998, 6, 351. [66] B. O. Brandsdal, A. O. Smalås, J. Åqvist, FEBS Lett. 2001, 499,

171.[67] T. Lonhienne, C. Gerday, G. Feller, Biochim. Biophys. Acta 2000,

1543, 1.

54

[68] M. Olufsen, A. O. Smalås, E. Moe, B. O. Brandsdal, J. Biol. Chem. 2005, 280, 18042.

[69] G. Wiegand, S. J. Remington, Annu. Rev. Biophys. Biophys. Chem. 1986, 15, 97.

[70] S. Remington, G. Wiegand, R. Huber, J. Mol. Biol. 1982, 158, 111. [71] R. J. M. Russell, J. M. C. Ferguson, D. W. Hough, M. J. Danson, G.

L. Taylor, Biochemistry 1997, 36, 9983. [72] I. Feierberg, J. Åqvist, Theor. Chem. Acc. 2002, 108, 71. [73] O. Donini, T. Darden, P. A. Kollman, J. Am. Chem. Soc. 2000, 122,

12270. [74] A. J. Mulholland, W. G. Richards, Proteins: Struct., Funct., Genet.

1997, 27, 9. [75] A. J. Mulholland, W. G. Richards, J. Phys. Chem. B 1998, 102,

6635. [76] W. Yang, D. G. Drueckhammer, J. Phys. Chem. B 2003, 107, 5986. [77] A. J. Mulholland, P. D. Lyne, M. Karplus, J. Am. Chem. Soc. 2000,

122, 534. [78] U. Gerike, M. J. Danson, N. J. Russell, D. W. Hough, Eur. J. Bio-

chem. 1997, 248, 49. [79] M. A. Arnott, R. A. Michael, C. R. Thompson, D. W. Hough, M. J.

Danson, J. Mol. Biol. 2000, 304, 657. [80] L. C. Kurz, G. Drysdale, M. Riley, M. A. Tomar, J. Chen, R. J. M.

Russell, M. J. Danson, Biochemistry 2000, 39, 2283. [81] A. Warshel, Proc. Natl. Acad. Sci. U. S. A. 1978, 75, 5250. [82] K. Ersmark, M. Nervall, E. Hamelink, L. K. Janka, J. C. Clemente,

B. M. Dunn, M. J. Blackman, B. Samuelsson, J. Åqvist, A. Hallberg, J. Med. Chem. 2005, 48, 6090.

[83] K. Ersmark, M. Nervall, H. Gutierrez-de-Teran, E. Hamelink, L. K. Janka, J. C. Clemente, B. M. Dunn, A. Gogoll, B. Samuelsson, J. Åqvist, A. Hallberg, Bioorg. Med. Chem. 2006, 14, 2197.

[84] H. Gutiérrez-de-Terán, M. Nervall, K. Ersmark, B. M. Dunn, A. Hallberg, J. Åqvist, Biochemistry 2006, 45, 10529.

[85] J. Liu, E. S. Istvan, I. Y. Gluzman, J. Gross, D. E. Goldberg, Proc.Natl. Acad. Sci. U. S. A. 2006, 103, 8840.

[86] H. G. Xiao, A. F. Sinkovits, B. C. Bryksa, M. Ogawa, R. Y. Yada, Protein Expr. Purif. 2006, 49, 88.

[87] N. Andreeva, P. Bogdanovich, I. Kashparov, M. Popov, M. Sten-gach, Proteins: Struct., Funct., Bioinf. 2004, 55, 705.

[88] E. Z. Eisenmesser, D. A. Bosco, M. Akke, D. Kern, Science 2002,295, 1520.

[89] W. W. Cleland, P. A. Frey, J. A. Gerlt, J. Biol. Chem. 1998, 273,25529.

[90] S. Hur, T. C. Bruice, J. Am. Chem. Soc. 2003, 125, 5964. [91] M. Cascella, C. Micheletti, U. Rothlisberger, P. Carloni, J. Am.

Chem. Soc. 2005, 127, 3734. [92] S. Piana, D. Bucher, P. Carloni, U. Rothlisberger, J. Phys. Chem. B

2004, 108, 11139.

55

[93] P. A. Fields, G. N. Somero, Proc. Natl. Acad. Sci. U. S. A. 1998, 95,11476.

[94] G. S. Bell, R. J. M. Russell, H. Connaris, D. W. Hough, M. J. Dan-son, G. L. Taylor, Eur. J. Biochem. 2002, 269, 6250.

[95] R. J. M. Russell, D. W. Hough, M. J. Danson, G. L. Taylor, Struc-ture 1994, 2, 1157.

[96] W. D. Cornell, P. Cieplak, C. I. Bayly, I. R. Gould, K. M. Merz, D. M. Ferguson, D. C. Spellmeyer, T. Fox, J. W. Caldwell, P. A. Koll-man, J. Am. Chem. Soc. 1995, 117, 5179.

[97] A. D. MacKerell, D. Bashford, M. Bellott, R. L. Dunbrack, J. D. Evanseck, M. J. Field, S. Fischer, J. Gao, H. Guo, S. Ha, D. Joseph-McCarthy, L. Kuchnir, K. Kuczera, F. T. K. Lau, C. Mattos, S. Michnick, T. Ngo, D. T. Nguyen, B. Prodhom, W. E. Reiher, B. Roux, M. Schlenkrich, J. C. Smith, R. Stote, J. Straub, M. Watanabe, J. Wiorkiewicz-Kuczera, D. Yin, M. Karplus, J. Phys. Chem. B 1998, 102, 3586.

[98] A. D. Mackerell, J. Wiorkiewiczkuczera, M. Karplus, J. Am. Chem. Soc. 1995, 117, 11946.

[99] W. L. Jorgensen, D. S. Maxwell, J. TiradoRives, J. Am. Chem. Soc. 1996, 118, 11225.

[100] W. F. van Gunsteren, H. J. C. Brendsen, Groningen Molecular Simulation (GROMOS) Library Manual, Nijenborgh, Groningen, The Netherlands, 1987.

[101] C. Oostenbrink, A. Villa, A. E. Mark, W. F. Van Gunsteren, J.Comput. Chem. 2004, 25, 1656.

[102] H. J. C. Berendsen, J. P. M. Postma, W. F. van Gunsteren, J. Her-mans, Interaction models for water in relation to protein hydration,Riedel, Dordrecht, The Netherlands, 1981.

[103] W. Jorgensen, J. Chandrasekhar, J. Madura, I. Rw, M. Klein, J.Chem. Phys. 1983, 79, 926.

[104] G. King, A. Warshel, J. Chem. Phys. 1989, 91, 3647. [105] J. Marelius, K. Kolmodin, I. Feierberg, J. Åqvist, J. Mol. Graph.

Model. 1998, 16, 213. [106] T. J. A. Ewing, I. D. Kuntz, J. Comput. Chem. 1997, 18, 1175. [107] M. Rarey, B. Kramer, T. Lengauer, G. Klebe, J. Mol. Biol. 1996,

261, 470. [108] R. A. Friesner, J. L. Banks, R. B. Murphy, T. A. Halgren, J. J.

Klicic, D. T. Mainz, M. P. Repasky, E. H. Knoll, M. Shelley, J. K. Perry, D. E. Shaw, P. Francis, P. S. Shenkin, J. Med. Chem. 2004,47, 1739.

[109] T. A. Halgren, R. B. Murphy, R. A. Friesner, H. S. Beard, L. L. Frye, W. T. Pollard, J. L. Banks, J. Med. Chem. 2004, 47, 1750.

[110] M. L. Verdonk, J. C. Cole, M. J. Hartshorn, C. W. Murray, R. D. Taylor, Proteins: Struct., Funct., Genet. 2003, 52, 609.

[111] M. D. Eldridge, C. W. Murray, T. R. Auton, G. V. Paolini, R. P. Mee, J. Comput. Aided Mol. Des. 1997, 11, 425.

[112] H. F. G. Velec, H. Gohlke, G. Klebe, J. Med. Chem. 2005, 48, 6296.

56

[113] R. X. Wang, L. H. Lai, S. M. Wang, J. Comput. Aided Mol. Des. 2002, 16, 11.

[114] G. Jones, P. Willett, R. C. Glen, A. R. Leach, R. Taylor, J. Mol. Biol. 1997, 267, 727.

[115] A. R. Leach, Molecular modelling - Principles and applications,Pearson Education Limited, Harlow, 2001.

[116] J. Marelius, M. Graffner-Nordberg, T. Hansson, A. Hallberg, J. Åqvist, J. Comput. Aided Drug Des. 1998, 12, 119.

[117] D. A. McQuarrie, Statistical mechanics, University Science Books, Sausalito, 2000.

[118] R. W. Zwanzig, J. Chem. Phys. 1954, 22, 1420. [119] B. O. Brandsdal, F. Österberg, M. Almlöf, I. Feierberg, V. B. Luzh-

kov, J. Åqvist, in Protein Simulations, Vol. 66, 2003, pp. 123. [120] P. A. Kollman, I. Massova, C. Reyes, B. Kuhn, S. H. Huo, L.

Chong, M. Lee, T. Lee, Y. Duan, W. Wang, O. Donini, P. Cieplak, J. Srinivasan, D. A. Case, T. E. Cheatham, Acc. Chem. Res. 2000,33, 889.

[121] F. S. Lee, Z. T. Chu, M. B. Bolger, A. Warshel, Protein Eng. 1992,5, 215.

[122] Y. Sham, Z. T. Chu, H. Tao, A. Warshel, Proteins 2000, 39, 393. [123] J. Åqvist, C. Medina, J. E. Samuelsson, Protein Eng. 1994, 7, 385. [124] J. Marelius, T. Hansson, J. Åqvist, Int. J. Quantum Chem. 1998, 69,

77.[125] J. Åqvist, A. Warshel, Chem. Rev. 1993, 93, 2523. [126] V. Luzhkov, J. Åqvist, J. Am. Chem. Soc. 1998, 120, 6131.

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 270

Editor: The Dean of the Faculty of Science and Technology

A doctoral dissertation from the Faculty of Science andTechnology, Uppsala University, is usually a summary of anumber of papers. A few copies of the complete dissertationare kept at major Swedish research libraries, while thesummary alone is distributed internationally through theseries Digital Comprehensive Summaries of UppsalaDissertations from the Faculty of Science and Technology.(Prior to January, 2005, the series was published under thetitle “Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-7468

ACTAUNIVERSITATISUPSALIENSISUPPSALA2007