Energetic Mapping of Transition State Analogue Interactions with Human and Plasmodium falciparum Purine Nucleoside Phosphorylases*♦

Human purine nucleoside phosphorylase (huPNP) is essential for human T-cell division by removing deoxyguanosine and preventing dGTP imbalance. Plasmodium falciparum expresses a distinct PNP (PfPNP) with a unique substrate specificity that includes 5′-methylthioinosine. The PfPNP functions both in purine salvage and in recycling purine groups from the polyamine synthetic pathway. Immucillin-H is an inhibitor of both huPNP and PfPNPs. It kills activated human T-cells and induces purine-less death in P. falciparum. Immucillin-H is a transition state analogue designed to mimic the early transition state of bovine PNP. The DADMe-Immucillins are second generation transition state analogues designed to match the fully dissociated transition states of huPNP and PfPNP. Immucillins, DADMe-Immucillins and related analogues are compared for their energetic interactions with human and P. falciparum PNPs. Immucillin-H and DADMe-Immucillin-H are 860 and 500 pm inhibitors against P. falciparum PNP but bind human PNP 15–35 times more tightly. This common pattern is a result of kcat for huPNP being 18-fold greater than kcat for PfPNP. This energetic binding difference between huPNP and PfPNP supports the kchem/kcat binding argument for transition state analogues. Preferential PfPNP inhibition is gained in the Immucillins by 5′-methylthio substitution which exploits the unique substrate specificity of PfPNP. Human PNP achieves part of its catalytic potential from 5′-OH neighboring group participation. When PfPNP acts on 5′-methylthioinosine, this interaction is not possible. Compensation for the 5′-OH effect in the P. falciparum enzyme is provided by improved leaving group interactions with Asp206 as a general acid compared with Asn at this position in huPNP. Specific atomic modifications in the transition state analogues cause disproportionate binding differences between huPNP and PfPNPs and pinpoint energetic binding differences despite similar transition states.

Inhibition of human purine nucleoside phosphorylase (huPNP) 1 by transition state analogue inhibitors shows promise for the control of T-cell cancers and autoimmune diseases (1)(2)(3). The combined inhibition of human and P. falciparum PNPs kills parasites cultured in human erythrocytes by purine-less death (4,5). Immucillin-H (ImmH) is the first transition state analogue inhibitor developed in this class and shows a 56 pM K d for huPNP (6). It has recently entered phase II clinical trials against T-and B-cell cancers (7). ImmH was designed from the early ribooxacarbenium ion transition state structure of bovine PNP (8). More recently, the transition state structures have been solved for huPNP and PfPNP, which are both characterized by symmetric near fully dissociated ribooxacarbenium ion transition states (9).
The second generation transition state analogue inhibitors were synthesized to mimic these dissociated transition states and include DADMe-Immucillin-H and DADMe-Immucillin-G. These are extraordinary inhibitors of huPNP with K d values of 16 and 7 pM, respectively. DADMe-Immucillin-H (DADMe-ImmH) has been called the "ultimate inhibitor" for mammalian PNP because of its action in a mouse model of PNP inhibition. A single oral dose of 27 g/mouse (0.1 mol) causes complete inhibition of blood PNP. Recovery of PNP activity in the blood is closely linked to the generation of new erythrocytes rather than inhibitor loss (10). Thus, administration of a single oral dose inhibits the target enzyme for the lifetime of erythrocytes, and additional inhibitory potential would have little additional physiologic effect.
Despite the similar transition state structures for huPNP and PfPNP, the substrate specificity, catalytic site structure, and subunit structures of PfPNP are distinct from huPNP (11). Here we investigate catalytic site differences between huPNP and PfPNP using an expanded family of first and second generation Immucillins.
Plasmodium falciparum cultured in human erythrocytes without added hypoxanthine are killed by ImmH due to purine starvation in these purine auxotrophs (4,5). The addition of exogenous hypoxanthine but not inosine rescues the parasites from killing by ImmH. This metabolic rescue pattern supports the Immucillin metabolic block at PNP (4). ImmH binds more tightly to huPNP than to PfPNP and inhibitors that bind selectively to PfPNP are important to establish if both huPNP and PfPNP must be blocked to induce purine-less death.
Immucillin and DADMe-Immucillin inhibitor families have been expanded to include the 5Ј-methylthio functional group specific for the P. falciparum enzyme (11). DADMe-ImmG and DADMe-ImmH are the tightest binding inhibitors for both huPNP and PfPNP. A methylene group is used in DADMe-ImmH to increase the distance between the ribooxacarbenium cationic mimic and the 9-deazahypoxanthine leaving group. These features closely resemble the transition states for both huPNP and PfPNP. DADMe-ImmH binds tighter to huPNP than ImmH; however, it binds less well to PfPNP, despite its similarity in ribooxacarbenium character to the PfPNP transition state. Differences in catalytic site properties can explain the difference in binding energy for transition state analogues of human and P. falciparum PNPs. The catalytic turnover number for huPNP (inosine) is 18-fold greater than for Pf-PNP. Binding of transition state analogue inhibitors is proportional to the catalytic rate enhancement imposed by the enzyme (12). Inhibitors capturing similar features of the transition state for inosine phosphorolysis are therefore expected to bind ϳ18-fold weaker to the malarial enzyme and this is observed in the inhibitor family. However, 5Ј-methylthioinosine (MTI) is a good substrate for PfPNP and a weaker substrate for human PNP (11). Thus, transition state mimics of MTI that contain a 5Ј-methylthio group are bound more tightly to PfPNP.
Human PNP achieves the transition state by a combination of ribooxacarbenium ion stabilization and purine leaving group activation. A mechanistic feature for ribooxacarbenium ion formation for huPNP involves crowding or overlap of the lone pair electrons from the 5Ј-hydroxyl to interact with the ribosyl 4Јring oxygen. This facilitates the electron "push" into the leaving group (13). The action of PfPNP on 5Ј-methylthioinosine precludes use of the 5Ј-oxygen lone pair mechanism. Immucillin-H binds to human PNP as the neutral molecule and in a second, slow onset tight binding step is protonated at 4Ј-N in the enzyme active site to form the cationic mimic of the transition state (14). This pattern is also involved in catalysis where neutral inosine binds and becomes the cationic ribooxacarbenium ion at the transition state. Recently, the x-ray crystal structure was solved for PfPNP with ImmH and MT-ImmH at the catalytic sites (11). Together with the transition state analogue inhibitor specificity shown here, it is apparent that the leaving group interactions for PfPNP are stronger than for huPNP as they must compensate for the loss of the 5Ј-hydroxyl lone pair interaction.
Preferential binding of transition state analogue inhibitors to PfPNP was observed only in the case of MT-Immucillin-H (Fig. 1). Interaction of other analogues with huPNP and PfPNP provides a geometric and electrostatic map for interactions at their catalytic sites. The specificity studies support two essential features for tight binding to both PNPs, elevated pK a of the leaving group, and the ability to form a cation to mimic the ribooxacarbenium ion transition state.

MATERIALS AND METHODS
Human PNP-Human PNP was expressed in a His 6 -N-terminal construct to facilitate purification and purified to homogeneity for these kinetic studies (9). The enzyme has also been expressed without Nterminal extension and the kinetic properties are unchanged by the N-terminal modification.
Inhibition Studies-Inhibitor dissociation constants for the phosphorolysis of inosine were based on initial and equilibrium reaction rate measurements with varied inhibitor concentrations (6,24). Reactions were started by adding huPNP (1.4 nM) or 1.0 g of PfPNP (32 nM) to reaction mixtures (25°C) containing 1 mM inosine in 50 mM KHPO 4 , pH 7.4, with xanthine oxidase at 60 milliunits/ml. Hypoxanthine formed by phosphorolysis of inosine was oxidized to uric acid and monitored spectrophotometrically at 293 nm (extinction coefficient for uric acid ⑀ 293 ϭ 12.9 mM Ϫ1 cm Ϫ1 ). The Michaelis constants for inosine were measured at the same phosphate concentration for huPNP and PfPNP. Enzyme concentration was adjusted to give absorbance changes not exceeding 1.0 during the time required to characterize initial and final slow onset inhibition equilibria. The large excess of substrate and continuous product depletion provided extended initial rate conditions. In most cases the inhibitor concentration was Ͼ10-fold greater than the enzyme concentration as required for simple analysis of two-state slow onset tight binding inhibition (24). In the cases of the most tightly bound inhibitors it was not possible to maintain this condition and corrections were made to compensate for the concentration of bound inhibitor (10). The inhibition constant K i describes the reversible equilibrium between enzyme and inhibitor for the initial inhibitor binding step. K i was determined by fitting the initial rates at different inhibitor concentration to the equation for competitive inhibition: i ϭ (k cat ϫ S)/(K m (1 ϩ I/ K i ) ϩ S), where i is initial reaction rate, k cat is the catalytic turnover number, K m is the Michaelis constant, K i is the dissociation constant of enzyme-inhibitor complex (EI), I is inhibitor concentration, and S is substrate concentration. The dissociation constant for the complex formed after slow onset equilibrium (K i *) was determined by ϭ (k cat ϫ S)/(K m (1 ϩ I/K i *) ϩ S), where is the steady state reaction rate and the other variables are the same as above.

RESULTS AND DISCUSSION
Inhibition of huPNP by Immucillins-ImmH, ImmG, and their 2Ј-deoxy analogues (2Ј-D-ImmH and 2Ј-D-ImmG) are the four most potent Immucillin inhibitors for huPNP with dissociation constants of 42 to 180 pM (Table I; Fig. 1). 3 Two 8-sub-stituted Immucillins, 8-aza-ImmH and 8-F-ImmH, are also pM inhibitors with dissociation constants of 180 and 390 pM, respectively, but the 8-Me group increases the dissociation constant to 20 nM, similar to N7 methylation, which gives a dissociation constant of 100 nM for 7-Me-ImmH. Immucillin-H molecules with 2Ј-alterations (2Ј,2Ј-difluoro, 2Ј-ara, 2Ј-methoxy) are still good inhibitors of with dissociation constants of 1.4 -5.9 nM. Thus, the catalytic site of huPNP is tolerant at the 2Ј-position, as expected for its substrate specificity to accept both inosine and guanosine and their 2Ј-deoxy analogues with approximately equal catalytic efficiency. 3Ј-D-ImmH gave a dissociation constant of 7.5 nM indicating that 3Ј-hydroxyl contacts are more important than those at the 2Ј-hydroxyl. Alterations in the exocyclic 6-carbonyl oxygen to give 6-MeO-ImmH or 6-S-ImmH increased the dissociation constants to 4.7 and 25 nM. Changes in the 5Ј-hydroxyl group causes large losses in affinity for huPNP, consistent with the catalytic mechanism involving the neighboring group participation of the lone pair of the 5Ј-hydroxyl in stabilizing the ribooxacarbenium ion (13). Thus, the 5Ј-F-ImmH, 5Ј-D-ImmH, 5Ј-PhT-ImmH, 5Ј-MT-ImmH, 5Ј-CONH 2 -ImmH, and 5Ј-COOH-ImmH increase K i values to 6.8 nM, 25 nM, 250 nM, 300 nM, Ͼ120 M, and Ͼ190 M, respectively. Binding of 1Ј-aza-inosine and 4Ј-N-ImmH to huPNP was not detected at 38 -240 M (Fig. 1). Addition of a propyl group to N4Ј or switching the ring N3 and C4 atoms decreased affinity to give dissociation constants of 410 nM and 1.9 M, respectively. The changes in ImmH chemical structure summarized in Fig. 1 cause up to a 5,700,000-fold (9.4 kcal/mol) change in binding affinity for huPNP.
An important feature of the specificity result ( Fig. 1) is that ImmH and ImmG, the transition state analogues originally designed for bovine PNP, are the tightest binding Immucillin inhibitors for both huPNP and bovine PNP (26). Second, there are six Immucillins with dissociation constants in the pM range. The pM inhibitors are modified only at the C8 or by being 2Ј-deoxy.
The pK a of N7 is an important determinant for transition state formation for mammalian PNPs. Modest perturbations in pK a values with little alteration in steric volume cause only modest changes in K i . Thus, 8-F-ImmH and 8-aza-ImmH have K i values of 390 and 180 pM, while the bulkier 8-Me-ImmH has K i increased to 20 nM. Likewise, 7-Me-ImmH has a K i of 100 nM. The ability of N7 substituted purine nucleosides to bind and to act as substrates of PNPs has been used as a rationale to propose that the H-bond that forms between the carbonyl of Asn 243 and N7 at the transition state is not important to catalysis. In contrast, N7 methylation generates a cationic site on the purine and serves to chemically activate the purine leaving group, making the H-bond at N7 unnecessary for leaving group departure. Crystallographic studies with huPNP have demonstrated that Asn 243 , the group in H-bond contact with N7, is mobile and can move to accommodate N7 substituents, a finding consistent with the modest perturbation in Immucillin binding by the N7-methyl substitution and the observation that 7-Me-inosine is a substrate for huPNP (27).
Purine nucleoside phosphorylases have no protein contacts to O4Ј of purine nucleosides or to N4Ј of Immucillins, permitting space in the catalytic site for the ribosyl group to migrate between the purine and the phosphate (13,28). Even the relatively bulky 4Ј-N-Pr-ImmH (K i ϭ 410 nM) binds better than substrate (K m ϳ 30 M) confirming the ability of huPNP to accept this volume, even though binding 7,300-fold more weakly than ImmH.
Transition state analogues require faithful representation of both geometric and electrostatic properties of the transition state. 4Ј-N-ImmH violates this principle by linking the purine to the 4Ј-imino nitrogen, a geometric change. No inhibition is detected at 38 M, indicating a million-fold loss in affinity. Similar consequences accompany electrostatic changes. Thus, DAD-1Ј,9-hypoxanthine and 4-aza-3-deaza-ImmH have low pK a values at N7, while ImmH and related analogues with 9-deazahypoxanthine rings have a pK a of above 10 for N7 (14). Without the increased basicity at N7, tight binding inhibition is not detected, resulting in up to 4.3 million-fold loss in binding affinity, despite the geometric similarity of these molecules to both substrate and transition state.
Inhibition of PfPNP by Immucillins-The k cat values for phosphorolysis of inosine by huPNP and PfPNP are 31 s Ϫ1 and 1.7 s Ϫ1 , respectively, making huPNP 18-fold more efficient than PfPNP. Catalytic efficiency is only an approximation of transition state stabilization for these PNPs, since chemistry is not completely rate-limiting for either enzyme (9). On the basis of transition state principles, it would be expected that transition state analogues of inosine would bind more weakly to PfPNP by the rate enhancement factor. This relationship is confirmed for the better Immucillin inhibitors (Fig. 1). Thus, ImmH, ImmG, 2Ј-d-ImmH 2Ј-d-ImmG, 8-F-ImmH, 8-aza-ImmH, and 3Ј-d-ImmH are relatively faithful mimics of an early ribooxacarbenium ion transition state and contain the essential recognition elements of the elevated pK a at N7, the imino cation at N4Ј, and the 5Ј-hydroxyl. These inhibitors bind better to the huPNP by factors of 15-, 21-, 16-, 47-, 23-, 72-, and 6-fold, respectively, in good agreement with the approximate 18-fold difference in transition state efficiency. Deviations from this relationship are seen in the analogues where the iminoribitol character, features of leaving group activation, or the 5Ј-hydroxyl modifications are made. Thus, 5Ј-MT-ImmH and 5Ј-PhT-ImmH bind 112-and 2-fold better to PfPNP than to human PNP. This reversal reflects the PfPNP substrate specificity for 5Ј-methylthioinosine, which is equal to inosine as a substrate for PfPNP (11). The methylthio group is found in a hydrophobic region of the catalytic site for PfPNP and this site also favors interaction with the larger phenyl group.
The ability of PfPNP to use 5Ј-methylthioinosine as a substrate precludes the lone pair interaction between the 5Ј-hydroxyl and the 4Ј-ring oxygen, an interaction proposed to be important for phosphorolysis of inosine and guanosine (29). Loss of this transition state stabilizing interaction may be associated with the decreased k cat for PfPNP. However, the PfPNP compensates for this loss with other interactions to increase catalytic efficiency. Strengthened interactions to N7, O6, and O 2 Ј at the transition state are all possibilities. Interaction at the 2Ј-hydroxyl is more important for the malarial enzyme, since 2Ј-MeO-ImmH and 2Ј,2Ј-diF-ImmH bind huPNP 140 and Ͼ3,300 times more tightly than PfPNP. Increased H-bond donation from the enzyme to the 2Ј-oxygen would decrease electronic contribution into the carbocation and facilitate transition state formation.
Volume near the 4Ј-imino cation is more restricted at the catalytic site of PfPNP than for huPNP, since N4Ј-Pr-ImmH binding was not detected at 300 M with PfPNP but binds with a dissociation constant of 410 nM to huPNP. The confined geometry around the iminoribitol cation in PfPNP is also apparent with DAD-1Ј, 9-hypoxanthine and 4Ј-N-ImmH, since neither bind to PfPNP at concentrations of 240 -870 M. However, these inhibitors also fail to bind to huPNP.
Transition states of mammalian PNPs involve leaving group activation by N7 protonation or H-bonding at N7 via a carbonyl oxygen hydrogen bond to Asn 243 , in addition to interactions at is the equilibrium dissociation constant for huPNP, and the middle line (in red) is the equilibrium dissociation constant for PfPNP. Where the value is a K i *, the dissociation constant occurs following slow onset tight binding inhibition as described under "Materials and Methods." Where the value is a K i value, there is no slow onset, and a single dissociation constant is observed. Both K i * and K i values listed here are thermodynamic dissociation constants and can be directly compared. The bottom line is in blue to indicate tighter binding to huPNP and in red to indicate tighter binding to PfPNP. Note that only two inhibitors bind more tightly to PfPNP than to huPNP and are shown in bold. The value given in the bottom line is the ratio of dissociation constants. Atomic numbering used for all inhibitors is given for ImmH.
O6 and N1 (13,21). PfPNP has Asp 206 as a general acid at this position and acts as a more favorable H-bond partner to N7, promoting leaving group activation (11). Interaction between Asp 206 and N7 is explored with N7 methylation of ImmH in 7-Me-ImmH (Fig. 1). While 7-Me-ImmH is a 4.7 nM inhibitor of huPNP, no inhibition of PfPNP was detected at 240 M; thus, the discrimination factor is Ͼ51,000 against PfPNP. The N7methyl effect supports the proposal that more of the transition state energy in PfPNP arises from the leaving group interaction at N7 than in huPNP.
Chemical stability of the DADMe-Immucillins requires them to be 2Ј-deoxy. The physiological substrate for huPNP is 2Јdeoxyguanosine; thus, the 2Ј-deoxy feature of DADMe-ImmG increases its structural mimicry of the transition state with  (25). Some values reported here differ from the earlier report and reflect use of overexpressed PNPs and more extensive kinetic analysis with both huPNP and PfPNP. b ND, there is no slow onset observed, thus, only K i is measured from inhibition studies. c When no inhibition was observed at a given concentration, the K i limit was set at three times the inhibitor concentration that gave no inhibition.  (10). b ND, there is no slow onset observed; thus, only K i is measured from inhibition studies. c When no inhibition was observed at a given concentration, the K i limit was set at three times the inhibitor concentration that gave no inhibition.

TABLE II Initial (K i ) and equilibrium (K i *) dissociation constants of methylene-bridged immucillin analogues for huPNP and PfPNP
dGuo. Placement of the cation-generating pyrrolidine nitrogen at the 1Ј-position more accurately mimics the carbocation electrostatics of this dissociated S N 1-like transition state (9). Accordingly, DADMe-ImmG is the most tightly bound transition state analogue yet known for huPNP. The leaving group pK a value at N7 is also important to capture transition state binding energy. 8-Aza-DADMe-ImmH has a decreased pK a at N7 (9.6) compared with that of DADMe-ImmH, which is Ͼ10, based on the pK a for Immucillin-H (14). This difference causes a 125-fold decrease in binding affinity. The hydroxypyrrolidine ring-opened analogues (4Ј-OH-5Јnor-3Ј,4Ј-seco-DADMe-ImmH and 3Ј,4Ј-seco-DADMe-ImmH) retain the 9-deazahypoxanthine feature of DADMe-ImmH and bind with 1.3-120 nM affinity, a decrease in binding energy reflecting the increased entropic cost of freely rotating groups (Fig. 2). Additional volume and a potential for new H-bonds in the hydroxypyrrolidine are provided by the 4Ј-hydroxyl substituent in 4Ј-OH-DADMe-ImmH. This change decreases binding by 875-fold relative to DADMe-ImmH to 14 nM. The decrease in binding due to the 4Ј-hydroxyl substituent between 5Ј-MT-DADMe-ImmH and its 4Ј-hydroxy derivative is similar at a 840-fold decrease, emphasizing the importance of geometry around the ribooxacarbenium ion mimic (Fig. 2).
The methylene bridge that characterizes the DADMe-Immucillins adds both distance and angular geometric relationships between the deazapurine and the ribocation mimic. This geometry was explored with an ethyl bridge in DADEt-ImmH, a methylene bridge to N4Ј of iminoribitol in 4Ј,9-Me-ImmH, a link to C8 of the deazapurine in 1Ј,8-DADMe-ImmH, and a methylene spacer inserted into ImmH in 1Ј,9-Me-ImmH. These alterations decreased binding affinity relative to DADMe-ImmH, by factors of 29, 170, 410,000, and 16,000 respectively. 1Ј,9-Me-ImmA retains the high pK a at N7 and the iminoribitol group as the carbocation mimic, but the 6-amino group prevents interaction at the catalytic site and the K i value is Ͼ150 M.
Inhibition of PfPNP by DADMe-Immucillins-The altered catalytic capacity of PfPNP relative to huPNP predicts ϳ18fold decreased binding affinity of transition state analogues to the Pf enzyme (see above). The DADMe-Immucillins, in every example, bind more tightly to huPNP than PfPNP (Table II and Fig. 2). DADMe-ImmH, DADMe-ImmG, and MT-DADMe-ImmH are the most tightly bound inhibitors for PfPNP at 500, 860, and 900 pM, respectively, and are the only pM inhibitors for PfPNP in the DADMe-Immucillin family. The relative affinity for huPNP and PfPNP is close to the values expected from transition state stabilization with DADMe-ImmH giving a 31-fold preference for huPNP. However, DADMe-ImmG shows a 130-fold preference for huPNP, reflecting the catalytic role of huPNP for dGuo phosphorolysis as compared with the physiological roles for PfPNP of forming hypoxanthine from inosine and 5Ј-methylthioinosine (5,11). For both physiological functions of PfPNP the substrate molecules contain a 2Ј-hydroxyl group. Other analogues, including 5Ј-d-5Ј-Me-DADMe-ImmH, 8-aza-DADMe-ImmH, and 1Ј,8-DADMe-ImmH differ only by 1.3-14-fold in affinity between huPNP and PfPNP.
Methylthio-Immucillin-H and Methylthio-DADMe-ImmH-A surprise in the inhibition pattern of DADMe-Immucillins is 5Ј-MT-DADMe-ImmH with its 13-fold preference for huPNP. In the Immucillin inhibitors, 5Ј-MT-ImmH binds 112-fold better to PfPNP than to huPNP and this is attributed to the 5Ј-methylthioinosine specificity for the PfPNP, a metabolite not found in human metabolism (5,11). Since DADMe-Immucillins are closer mimics of the highly dissociated transition states of huPNP and PfPNP, it was hypothesized that the 5Ј-methylthio group would also convey specificity for PfPNP in the DADMe-Immucillin series. Its failure to do so suggests that the 2Јhydroxyl group is in a cooperative binding interaction with the 5Ј-methylthio group and/or other determinants of the transition state. Cooperative binding interactions between individual groups of transition state analogues have been documented with bovine PNP (26). Transition state analysis for the arsenolysis reaction catalyzed by bovine PNP indicated significant bond order (0.38) to the leaving group without participation of the arsenate nucleophile (8). Loss of any feature critical to the transition state interaction weakens direct interactions with the catalytic site and also weakens nearby interactions related to the transition state (26). In the case of PfPNP, the 5Ј-methylthioinosine substrate and MT-ImmH have a 2Ј-hydroxyl group that is not present in 5Ј-MT-DADMe-ImmH. Thus, the loss of the 2Ј-hydroxyl for PfPNP is more significant than the replacement of the 5Ј-hydroxyl with 5Ј-thiomethyl for huPNP.
Energetics of Immucillin Binding to huPNP and PfPNP-The binding affinity for Immucillin transition state analogues of huPNP and PfPNP is compared with the affinity for ImmH in Fig. 3 (upper panel). Atomic substitutions at all positions except the 2-amino (ImmG) decrease binding affinity and therefore have a positive ⌬⌬G relative to ImmH binding with huPNP and PfPNP. Interesting atomic substitutions are those with large differentials in binding for the PNP isozymes. One of these is 8-Me-ImmH, which decreases the binding energy by 3.5 kcal/mol compared with ImmH for huPNP but 7.0 kcal/mol for PfPNP. This change is possible either through perturbation of the pK a at N7 or by a steric clash in the PfPNP that is not present in huPNP. The 8-F-ImmH answers this question, since the fluorine substitution also alters the pK a at N7 but does not alter the steric volume. The 8-F-ImmH decreases binding to both huPNP and PfPNP by the same amount. Therefore the large discrimination seen for the 8-methyl group indicates that the catalytic site of the PfPNP is more highly constrained The arrows indicate atomic substitutions at each position. For example, 3 OCH 3 and 3 S are hydroxymethyl and sulfur replacements for O6 of ImmH. The values adjacent to the atomic substitutions are the Gibbs free energy changes (kcal/mol) in binding energy caused by atomic substitution at that position relative to binding energy for ImmH. Energetic differences are calculated according to the expression: ⌬⌬G ϭ ϪRT ln (K d ImmH)/(K d Imm analogue), where R is the gas constant, T is absolute temperature, and the value of RT ϭ is 602.1 cal/mol. These ⌬⌬G energies compare ImmH and other analogues binding huPNP (values in blue) and ImmH and other analogues binding to PfPNP (in red). When the blue and red values are equal, a substitution at this position has the same effect on inhibitor binding for huPNP and PfPNP. When the numbers are different, atomic substitution at the indicated positions influence binding to one enzyme more than the other. An example is the CH 3 S-substitution at C5Ј where binding decreased by 5.2 kcal/mol for huPNP but only 0.7 kcal/mol for PfPNP, both relative to ImmH binding. The lower panel has the same energetic meanings for the DADMe-Immucillin-and methylene-bridged analogues. In the lower structure, seco indicates deletion of the 3Ј-4Ј covalent bond of DADMe-ImmH. around this position than huPNP. Another substitution of interest is in the 5Ј-methylthio group of 5Ј-MT-ImmH. The Pf-PNP uses 5Ј-MT-inosine as a physiological substrate and has a hydrophobic pocket to accommodate the methylthio group (11). Placing a methylthio group in the 5Ј-position for huPNP decreases affinity relative to ImmH by 5.2 kcal/mol but for PfPNP the binding affinity changes only by 0.7 kcal/mol. This substitution confers high specificity for 5Ј-MT-ImmH on the PfPNP (11). Recently, this feature has been used to probe the relative contributions of erythrocyte PNP and PfPNP to purine salvage in P. falciparum cultured in human erythrocytes (4). The third feature to give isozyme discrimination is substitution of the difluoro at the 2ЈC of the iminoribitol of ImmH (2Ј,2Ј-diF-ImmH). Human PNP experiences a decrease of 1.9 kcal/mol for this substitution relative to a hydroxyl at this position with ImmH, but PfPNP experiences a loss in binding energy of Ͼ5.9 kcal/mol. In contrast, huPNP and PfPNP are both tolerant of 2Ј-deoxy, 2Ј-ara, and 2Ј-OCH 3 substitutions, since all of these changes affect both enzymes with similar changes in ⌬⌬G. Thus, the electron-rich difluoro substitution reflects an unfavorable electrostatic interaction or induces an unfavorable iminoribitol ring pucker for PfPNP relative to huPNP. Although huPNP and PfPNP bind some Immucillins with large differences, for many other Immucillins, binding differences are unexpectedly small. For example, a catalytic site feature that most distinguishes huPNP from PfPNP is the hydrophobic cavity that accommodates the 5Ј-methylthio group, specifically for PfPNP. Substitution of a phenylthio group (5Ј-PhT-ImmH) at this position alters binding of both huPNP and PfPNP by a similar energy of ϳ5 kcal/mol.
Energetics of DADMe-Immucillin Binding to huPNP and PfPNP-The binding of DADMe-ImmH to huPNP and PfPNP exhibits dissociation constants of 16 and 500 pM, respectively. DADMe-ImmH is the most powerful inhibitor known for Pf-PNP and the second most powerful for huPNP. The rationale for this tight binding is the close mimicry of DADMe-ImmH to the fully dissociated transition states for both enzymes. Transition states for both huPNP and PfPNP have N-ribosidic bond (N9 to C1Ј) distances of ϳ3.0 Å compared with 1.5 Å in the substrate molecules (9). Introduction of the methylene bridge in the DADMe-Immucillins increases this distance to 2.5 Å (C9 to N1Ј) compared with a distance of 1.4 Å in the Immucillins (C9 to C1Ј). The deazapurine leaving group in the DADMe-Immucillins retains the high pK a at N7, and the pyrrole nitrogen cation at the 1Ј-position closely mimics the electrostatics of the transition state. DADMe-ImmG binds Ϫ0.5 kcal/mol more tightly than DADMe-ImmH to huPNP but binds 0.3 kcal/mol less well to PfPNP. While these are relatively small differences, they reflect the physiological roles of the enzymes. In humans PNP catabolizes dGuo and in P. falciparum PNP forms hypoxanthine from inosine and 5Ј-methylthioinos ine. The 2.5-Å distance separating leaving group and oxacarbenium ion mimics in DADMe-ImmH does not exactly match the distance or the geometry of these transition states. Insertion of an ethylene bridge increases the distance to 3.5 Å, greater than the 3.0 Å of the transition state, and also introduces additional conformational flexibility. DADEt-ImmH binding is decreased by 2.0 kcal/mol for huPNP but only 0.1 kcal/mol for the PfPNP relative to DADMe-ImmH. These changes suggest that the transition state for huPNP is more closely matched by the 2.5 Å in DADMe-ImmH than the 3.5-Å bond separation in DADEt-ImmH and that the PfPNP transition state has bond separation between 3.0 and 3.5 Å. These bond distances are consistent with kinetic isotope effect measurements, since leaving group distances Ͼ3.0 Å are not accurately predicted by KIE. Surpris-ingly, the 5Ј-methylthio and 5Ј-propylthio substitutions in the DADMe-Immucillins do not show strong preferences for the PfPNP. Transition state interactions at the catalytic sites of PNP are known to be cooperative. Since the DADMe-Immucillins are required to be 2Ј-deoxy for chemical stability, it is possible that cooperative binding occurs between the 2Ј-hydroxyl and the 5Ј-methylthio groups, and without the 2Ј-interaction, the methylthio group does not interact optimally with its hydrophobic site. This interpretation remains speculative pending structural analysis of these complexes.
Geometric links between the ribosyl and purine mimics were changed to introduce a methylene bridge between N4Ј and C9 in ImmH (4Ј,9-N-Me-ImmH), and this decreased binding by 3-4 kcal/mol for huPNP and PfPNP. Another altered link in 1Ј,8-DADMe-ImmH decreased binding by 7.8 and 6.8 kcal/mol, destroying the transition state analogue binding energy. Indeed, this analogue binds to PNPs with affinity similar to substrate molecules.
Summary and Conclusions-The Immucillin transition state analogues of human and P. falciparum PNPs give K m /K i * values up to 5,400,000 (10). Systematic atomic substitutions of these designed inhibitors only weaken the binding, supporting the proposal that these molecules are highly optimized to capture transition state binding energy. Although only a fraction of the transition state binding energy is captured, imperfect mimicry of transition state structure is a necessary feature of attempting to mimic unstable transition states with chemically stable analogues.